<?xml version="1.0" encoding="utf-8"?>
  <rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>Science</title>
      <link>https://blog--3ds--com.apsulis.fr/topics/science/feed.xml</link>
      <description>Science</description>
      <lastBuildDate>Thu, 05 Mar 2026 16:10:06 GMT</lastBuildDate>
      <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
      <generator>3DExperience Works</generator>
      <atom:link href="https://blog--3ds--com.apsulis.fr/topics/science/feed.xml" rel="self" type="application/rss+xml"/>

      <item>
      <title>
      <![CDATA[ How can we make biomanufacturing for life sciences and healthcare more sustainable? ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/industries/life-sciences-healthcare/sustainability-in-biomanufacturing/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/274611</guid>
      <pubDate>Wed, 04 Dec 2024 16:02:30 GMT</pubDate>
      <description>
      <![CDATA[ Learn how digitalization drives sustainability in biomanufacturing and helps companies create new treatments that don’t cost the Eart. 
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
When a team of Canadian researchers discovered how to isolate, extract, purify and administer insulin in the 1920s, they transformed the treatment of diabetes. Their use of this naturally occurring hormone to create medicine also kick-started the development of biomanufacturing in life sciences and healthcare.



Skip forward a century and biologics – biomanufactured drugs made from natural resources like microorganisms and cells – offer a vast range of personalized and targeted treatments. This is the sector that, in 2020, collaborated to develop, produce and distribute effective vaccines to combat COVID-19 in a matter of months – a process that would normally take years.



Today, biomanufacturers are innovating at pace, and growing their production, supply and distribution capacity so they can meet unmet medical needs and an increasing demand from an aging global population. But, currently Life Sciences and Healthcare therapy development and production activities generate 4.4% of global emissions, this means these companies cannot continue on this path – they must produce more with less resources.



Pressure on sustainability performance is coming from all sides. Measures that were once secondary concerns – like reducing waste, energy consumption and emissions – are now a top priority for business leaders. Companies face intense scrutiny from investors and customers when it comes to advancing sustainability in biomanufacturing while reducing costs. And they must also prove their performance against increasingly stringent regulations.



A closer look at biologics



Biologics excel in targeting the proteins, cells or pathways involved in a disease’s progression. Examples include monoclonal antibodies, which are designed to neutralize harmful proteins in the body. Then there are cytokines, which modify immune system responses to fight off or suppress a specific condition. Meanwhile vaccines prevent diseases from developing by stimulating the immune system to fight certain pathogens.



This targeting ability makes biologics effective for treating chronic diseases including autoimmune disorders, cancer, diabetes, inflammatory bowel disease and rheumatoid arthritis. They are also less likely to cause side effects than conventional drugs, which act more broadly across the body.



However, the size and complexity of the molecules involved means that biologics are more complex, time-consuming and costly to manufacture than conventional drugs, which are made from smaller, simpler molecules. The procedure typically begins when a gene, which encodes the desired protein, is transferred into “production cells” such as E. coli bacteria. These cells are then cultivated in a bioreactor under carefully controlled conditions such as temperature, pH and nutrient supply. When the cells are ready, they are processed to release the protein. This is then refined to make sure it meets the purity, potency and safety standards required.



These processes use significant natural resources and energy while producing emissions and waste. But biomanufacturers need to be sure that any measures they take to improve sustainability will not adversely affect the quality of their product, but must be also cognizant of the waste they create. To minimize that risk, sustainability must be designed into these delicate procedures, not bolted afterwards.



Where can biomanufacturers make sustainability gains?



To make meaningful sustainability gains, biomanufacturers must adopt a new paradigm and factor in sustainability concerns from the very beginning of drug development.&nbsp; From the ideation of the drug to the biologic process to facility development on through delivery, each stage in the development process can include sustainability indicators that impact the whole.



Let’s look at a few:




Product development. On average, 80% of any product’s environmental impact is made during the design phase. If biologic developers identify which manufacturing methods will yield the best results for their recipe and the environment, they can make a big dent in that figure.



Supply chain and logistics. These activities typically account for 60-80% of a pharmaceutical company’s greenhouse gas emissions. Those percentages come down when key performance indicators (KPIs) for sustainability are factored into supply chain design from the start.



Continual process improvement. By monitoring environmental performance alongside critical parameters – such as temperature, which is crucial in biologic production – it’s possible to identify small changes that will improve on the process &nbsp;without risking the product’s quality.




What challenges are slowing progress toward sustainability in biomanufacturing?



Knowing where to look is one thing. Finding the information can be quite another, especially since pharmaceutical companies tend to have volumes of scientific data across separate systems and departments. This disjointed view makes it difficult to build a common sustainability strategy across the organization or the product lifecycle. As a result, measures often focus on individual business units, which limits their success.



Factoring in third-party providers such as suppliers and logistics organizations throws up more obstacles. When decision-makers don’t have all the information available in one place, they can’t make informed decisions that balance sustainability alongside KPIs like time and cost.



And what about those opportunities for improvement that emerge during manufacturing? Without a holistic view, based on all the data involved, it’s impossible to know where to look, or what effect any decisions will have further down the line.



Digitalization can drive sustainability in biomanufacturing



Unifying data in one place – the 3DEXPERIENCE platform – has helped Dassault Systèmes customers across Life Sciences &amp; Healthcare and other industries to put sustainability performance in the context of other critical KPIs. The platform allows the creation of virtual twins – holistic digital models of real-world products, systems and processes. By using a virtual twin to structure and contextualize data it allows to see how different KPIs relate to each other across the product lifecycle, biomanufacturers can target new measures wherever they are needed,



It’s important for biomanufacturers to target areas where the biggest sustainability gains can be achieved. They can use a recipe lifecycle management solution, for instance, to see how different processes all along the product lifecycle – such as an energy-intensive or high-emission manufacturing method or some aspects of the supply chain– will affect a product’s environmental impact. Then they can make the appropriate changes to bring those figures down.



Solutions are also available to support the optimization of supply and demand planning. For example, a built-in algorithm can be used to show supply chain planners which suppliers perform best across KPIs like sustainability, reliability and cost, so they can choose the one that strikes the best balance between all those priorities.



Once the biologic enters production, monitoring processes and equipment in real time is essential to identify ongoing opportunities for improvement. This is what the &nbsp;Continued Process Verification toolset does. Its algorithm helps companies understand the different patterns at play in areas like energy consumption and emissions, as well as production parameters – so users can see where to make changes that will improve sustainability while staying within production parameters.



Sustaining the journey



There has always been a sound business incentive for improving sustainability. Minimizing waste and reducing energy consumption often go hand in hand with increased efficiency, as well as potentially attracting more customers and investors. What’s been missing until recently is the joined-up view that shows how and where sustainability can be optimized in balance with performance and quality KPIs across the product lifecycle.



Technologies like those discussed above are bringing that view to life, supported by powerful algorithms. We’re just starting to see how artificial intelligence, with its complex data analysis capabilities, can play a pivotal role in enabling biomanufacturers to thread sustainability across their business. In this intensely innovative industry, it will be exciting to see how companies work to achieve that.



Check out our new whitepaper, &#8220;Innovating for a More Sustinable Future: Leveraging Greener Practices in Biomanufacturing&#8221; for an even deeper dive into sustainability in biomanufacturing, including how advanced  digital tools like AI and virtual twins allow companies to improve efficiency and reduce emissions, waste, energy, and water usage while ensuring quality and safety.




Download the whitepaper today





 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ The Shape of Water ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/the-shape-of-water/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/273202</guid>
      <pubDate>Fri, 22 Nov 2024 01:16:25 GMT</pubDate>
      <description>
      <![CDATA[ Discover how BIOVIA Solvation Chemistry and COSMOtherm uncover the hydrophobic effect of water, a phenomenon fundamental to life and industry.
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Understanding the Hydrophobic Effect



Although it seems quite abstract, the interactions of molecules and atoms on a microscopic level often have profound effects on the macroscopic properties of materials and life itself. BIOVIA Solvation Chemistry provides the scientific link to understand these connections between microscopic, molecular interactions and industry-relevant experimental properties of liquids such as solubilities, vapor pressures, partition coefficients and many more. Exemplarily, in this blog post, we want to discuss the anomalous hydrophobic effect of water on a molecular level and its implications on industrial applications and life on Earth in general.



The phenomenon known as the hydrophobic effect, originating from the Greek words ύδωρ (ydor, water) and φόβος (phobos, fear), describes how molecules that “fear” water, tend to come together to avoid interacting with it. This effect explains, why oil and water do not mix but form separate phases, or why some compounds are better soluble in water than others.



It is not just dry textbook knowledge, but a concept fundamental to the existence of life on our planet, serving as a key driver behind biological processes such as the formation of cells and the folding of proteins in active or inactive structures. Proteins are essential building blocks of our body, consisting of chains of amino acids. In many cases, proteins only function properly when they are in a certain folded state, a state that is typically achieved within a specific range of temperatures. Among other factors, increased temperatures can lead to the denaturation of proteins, rendering them inactive. This process can be easily observed when e.g. boiling an egg, where egg albumen turns into a white, opaque substance upon denaturation of proteins (mostly ovalbumin) above 60°C. Also at lower temperatures, proteins can undergo reversible unfolding, a process known as cold denaturation. This behavior is directly linked to the temperature sensitivity of the hydrophobic effect and in order to understand it, we need to investigate the hydrophobicity of water itself.



The Surprising Hydrophobicity of Water



It might sound surprising, but water itself can be hydrophobic, at least to a certain degree.



Water molecules can be arranged in specific shapes or clusters. Here, molecules are connected by, what chemists call, a hydrogen bond. In water, this special type of bond occurs between a hydrogen atom and the oxygen atom of another water molecule. Since water has two hydrogen atoms and each oxygen atom can accommodate two hydrogen-bonds, complex networks can be formed, that stabilize different shapes of water clusters. Interestingly, the surface of these clusters has different properties than the surface of individual water molecules.



Relative screening charge density profile of the surface of a single water molecule (blue, structure top left) and a cluster of 20 molecules (orange, structure top right). Clearly the peak of charge neutral area in the center of the orange curve is visible, that is absent for isolated water molecules.







The picture shows the charge density surface of water and a cluster of water molecules connected by hydrogen bonds, easily computed with BIOVIA Turbomole Blue and red areas denote surface areas with large positive or negative screening charge densities, whereas green areas denote small screening charge densities close to zero. Generally, as a direct consequence of Coulomb’s law, opposing charges attract each other. Therefore, compounds with large positive or negative screening charge densities, prefer to be in contact with compounds of matching opposing screening charge densities. Compounds with a lot of nonpolar surface area (green) prefer other nonpolar compounds. As can be seen, water itself has quite a lot of positive and negative screening charge densities (blue, red). In contrast to this, the cluster structure, has more neutral area with small screening charge densities. For this reason, these clusters partly behave like a nonpolar hydrophobic substance.



Therefore, at low temperatures, when these clusters become more stable, surface properties of water change. Water itself becomes increasingly hydrophobic and in turn, hydrophobic molecules become more soluble in water. When the temperature increases, the clusters break apart and the solubility of hydrophobic molecules decreases. At even higher temperatures, other thermodynamic effects increase the solubility again, leading to a minimum in solubility, that is typically found somewhere in the range of 20 to 80 °C, i.e. near room and body temperature.



The accurate assessment of this temperature dependence is of importance in many applications, ranging from the solubility of additives in oil processing or carbon capture applications, to computation of partition coefficients of active pharmaceutical ingredients.



Advancing Innovation with BIOVIA COSMOtherm



BIOVIA COSMOtherm allows to simulate the temperature dependent hydrophobic effect of water in an efficient and accurate way, as is shown in a recent publication by M. P. Andersson and M. Richter



Aqueous solubilities of hexanol and benzaldehyde over a wide temperature range are shown. Experimental solubilities (blue curves) show a minimum around 320 K (hexanol) and 290 K (benzaldehyde) due to the hydrophobic effect of water. COSMOtherm FINE 2023 is able to recover this minima with high accuracy (325 K for hexanol and 290 K for benzaldehyde).







The picture shows the temperature dependent solubility of hexanol and benzaldehyde in water. Since both compounds are rather hydrophobic, the solubility is generally quite small, but simulation and experiment clearly show a minimum of solubility around 290 K (17 °C) for benzaldehyde and 320 K (47 °C) for hexanol. With BIOVIA COSMOtherm the temperature dependent solubility of any compound in water can be easily assessed out-of-the-box, including anomalies like the hydrophobic effect of water. This ability allows performing large-scale in-silico screenings of molecular properties with high accuracy that complement experimental efforts, speed-up innovation and reduce time-to-market for our customers.  



Learn more about BIOVIA COSMO RS.







References




Andersson, M.P. &amp; Richter, M. (2024). Comment on: The shape of water &#8211; how cluster formation explains the hydrophobic effect. J. Mol. Liq., 409, 125465 https://doi.org/10.1016/j.molliq.2024.125465








Interested in staying up to date on all the latest news of BIOVIA?




Subscribe to our newsletter





 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Interpretable Machine Learning in Pipeline Pilot ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/interpretable-machine-learning-in-pipeline-pilot/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/272613</guid>
      <pubDate>Fri, 15 Nov 2024 17:21:04 GMT</pubDate>
      <description>
      <![CDATA[ Explore how interpretable machine learning models like SISSO are advancing scientific insight in chemistry
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Introduction



The majority of machine learning algorithms applied in chemistry and biology are black box models used to make predictions on given target properties.1–3



The model receives input features and generates an output but the inner workings of how the model arrived at the output is unknown or extremely difficult to understand due to the complexity of the models. Therefore, extracting meaningful scientific insights from these models has proven to be a challenge.1,4



Interpretable ML models that offer predictive capabilities combined with interpretable physical equations are gaining traction in many areas of science.1,2,5,6



The goal here is to have what is termed a glass-box model where simple physical equations relate the input features to the target properties. In this way relationships in the data can be understood and improved scientific insight can be gained from the model.







Figure 1: Schematic of Black-box models and Interpretable glass-box models



Sure Independence Screening and Sparsifying Operator – SISSO



Among the methods developed for interpretable machine learning the Sure Independence Screening and Sparsifying Operator – SISSO methodology has been widely applied in heterogeneous catalysis and organic chemistry.7–11 SISSO is part of the symbolic regression class of models and can be used to find mathematical functions to predict the target property.



In simple terms SISSO consists of 2 parts:




Creation of a large feature space by combining the feature columns or descriptors with user selected operators (e.g., multiplication, division, ln, sqrt etc.).



Using sure-independence screening (SIS) to select the descriptors with highest correlation to the target property. Finally applying regularization (ℓ0) to select low-dimension linear models with the lowest error.




With this approach, the aim is to use SISSO to find interpretable equations, that make scientific sense, from a range of input features. The input columns or descriptors can be experimental and/or those obtained from molecular modelling studies, including those conducted in BIOVIA Materials Studio® or BIOVIA Turbomole®.







Figure 2: Example of the types of inputs that have been used with SISSO



The original SISSO code was implemented in FORTRAN12 and does not contain direct Python support. However, a newer C++ implementation (SISSO++) has been released by the NOMAD Laboratory which has native Python integration.13,14



Let’s say I wanted to apply the SISSO algorithm to some chemistry datasets in order to expand my scientific insight. How could I go about deploying this ML method as part of my data science pipelines?



The answer is to use BIOVIA Pipeline Pilot15 &nbsp;to wrap the Python code and extend access to this glass box models.



SISSO++ integration with Pipeline Pilot



Using the strong integration between BIOVIA Pipeline Pilot and Python there are a range of options one can take to incorporate Python code into existing data pipelines.16



In this example we are going to use the Jupyter Notebook components to handle the Python portion and use the native PLP components to read, write and clean data ready for input.



We will take two sets of data, one published by bp17 and one published by Sigman and co-workers.10 The bp data covers the use of benzaldehyde promoters in H-ZSM-5 dehydration of methanol to dimethyl ether (DME) and the Sigman data a diastereoselective Rh catalysed C-H insertion.



Both datasets are small (22 rows and 84 rows) by the standards of most AI methods but reflect the realistic acquisition of small high-quality datasets typically found in industry and academia.



Inside the Python Jupyter Notebook component, SISSO++ can be set up by selecting the operators, target column and the desired train/test split. In addition, hyperparameters can be set and the calculation type can be toggled between regression and classification.



We apply the model to the bp dataset where the target property is DME STY (space time yield – a measure of catalytic performance) and the 10 descriptor columns are density functional theory (DFT) derived features for the organic promoter aldehydes (other reaction parameters are kept constant).



We obtain an interpretable equation that provides scientific insight and outputs are displayed using the Pipeline Pilot reporting components. The SISSO++ model is comparable to the reported model and makes chemical sense as it relates steric and electronic features of the aldehyde promoter to catalytic performance.







Figure 3: Output of SISSO++ regression model for bp dataset run through BIOVIA Pipeline Pilot



One potential limitation of the SISSO++ code is that it can become computationally expensive with datasets that contain a large number of features. To that end, the Materials AI team at BIOVIA along with Felix Hanke (formerly at BIOVIA) developed a BIOVIA Pipeline Pilot native version of SISSO++ for regression problems.



Native SISSO++ in Pipeline Pilot



By using the parallelisation and simplicity of BIOVIA Pipeline Pilot we can increase the speed of finding interpretable equations for scientific datasets and run the models without any coding expertise.



The new protocol applies the same SISSO++ methodology whereby a vast number of features are generated and then parsed to give the best performing equations, but this is performed in a different way inside Pipeline Pilot.







Figure 4: Native SISSO protocol in BIOVIA Pipeline Pilot



The resulting output is comparable to the SISSO++ Python package but simplifies usage for the scientist as you do not need to interact with any code. In fact, the protocol can be run through the Pipeline Pilot Web Port with users selecting parameters through drop-down menus making it ideal for scientists with no coding experience.



In this example we show the output for the dataset from Sigman and co-workers10 where the target is ΔΔG‡ (a measure of diastereoselectivity) and there are 19 DFT derived chemical descriptors. Again, we obtain an interpretable equation for the data that is comparable to the reported model which related steric and electronic properties of the catalyst/ligand to the diastereoselectivity.



Due to the ability of Pipeline Pilot to handle large amounts of data effectively obtaining models with larger datasets (&gt;50 billion generated features) is also possible.







Figure 5: Example of BIOVIA Pipeline Pilot Web Port being used to run native SISSO algorithm.



Conclusion



The simple integration of Python in BIOVIA Pipeline Pilot enables us to incorporate SISSO++ and other Python packages easily into new and existing data pipelines.



We can also make use of the flexibility and speed of BIOVIA Pipeline Pilot to incorporate new methods for interpretable machine learning into data science workflows. In this way, BIOVIA Pipeline Pilot can be used to help scientists gain meaningful scientific insights from predictive models. With BIOVIA Pipeline Pilot these types of models can be deployed in a low/no-code environment to aid understanding and further innovation in scientific challenges.



References



(1)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Esterhuizen, J. A.; Goldsmith, B. R.; Linic, S. Interpretable Machine Learning for Knowledge Generation in Heterogeneous Catalysis. Nat Catal 2022, 5.



(2)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Azodi, C. B.; Tang, J.; Shiu, S.-H. Opening the Black Box: Interpretable Machine Learning for Geneticists. Trends in Genetics 2020, 36 (6), 442–455.



(3)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Jiménez-Luna, J.; Grisoni, F.; Schneider, G. Drug Discovery with Explainable Artificial Intelligence. Nature Machine Intelligence. Nature Research October 1, 2020, pp 573–584.



(4)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Molnar, C. Interpretable Machine Learning. A Guide for Making Black Box Models Explainable; https://christophm.github.io/interpretable-ml-book/., 2019, accessed 04/11/2025



(5)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; La Cava, W. G.; Lee, P. C.; Ajmal, I.; Ding, X.; Solanki, P.; Cohen, J. B.; Moore, J. H.; Herman, D. S. A Flexible Symbolic Regression Method for Constructing Interpretable Clinical Prediction Models. NPJ Digit Med 2023, 6 (1).



(6)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Rudin, C. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nat Mach Intell 2019, 1 (5), 206–215.



(7)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Foppa, L.; Rüther, F.; Geske, M.; Koch, G.; Girgsdies, F.; Kube, P.; Carey, S. J.; Hävecker, M.; Timpe, O.; Tarasov, A. V.; Scheffler, M.; Rosowski, F.; Schlögl, R.; Trunschke, A. Data-Centric Heterogeneous Catalysis: Identifying Rules and Materials Genes of Alkane Selective Oxidation. J Am Chem Soc 2023, 145 (6), 3427–3442.



(8)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Miyazaki, R.; Belthle, K. S.; Tüysüz, H.; Foppa, L.; Scheffler, M. Materials Genes of CO 2 Hydrogenation on Supported Cobalt Catalysts: An Artificial Intelligence Approach Integrating Theoretical and Experimental Data. J Am Chem Soc 2024, 146 (8), 5433–5444.



(9)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Wang, J.; Xie, H.; Wang, Y.; Ouyang, R. Distilling Accurate Descriptors from Multi-Source Experimental Data for Discovering Highly Active Perovskite OER Catalysts. J Am Chem Soc 2023, 145 (20), 11457–11465.



(10)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Souza, L. W.; Miller, B. R.; Cammarota, R. C.; Lo, A.; Lopez, I.; Shiue, Y.-S.; Bergstrom, B. D.; Dishman, S. N.; Fettinger, J. C.; Sigman, M. S.; Shaw, J. T. Deconvoluting Nonlinear Catalyst–Substrate Effects in the Intramolecular Dirhodium-Catalyzed C–H Insertion of Donor/Donor Carbenes Using Data Science Tools. ACS Catal 2023, 104–115.



(11)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Park, J.; Oh, J.; Kim, J. S.; Shin, J. H.; Jeon, N.; Chang, H.; Yun, Y. Catalyst Discovery for Propane Dehydrogenation through Interpretable Machine Learning: Leveraging Laboratory-Scale Database and Atomic Properties. ACS Sustain Chem Eng 2024, 12 (28), 10376–10386.



(12)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Ouyang, R.; Curtarolo, S.; Ahmetcik, E.; Scheffler, M.; Ghiringhelli, L. M. SISSO: A Compressed-Sensing Method for Identifying the Best Low-Dimensional Descriptor in an Immensity of Offered Candidates. Phys Rev Mater 2018, 2 (8), 1–12.



(13)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Purcell, T. A. R.; Scheffler, M.; Carbogno, C.; Ghiringhelli, L. M. SISSO++: A C++ Implementation of the Sure-Independence Screening and Sparsifying Operator Approach. J Open Source Softw 2022, 7 (71), 3960.



(14)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Purcell, T. A. R.; Scheffler, M.; Ghiringhelli, L. M. Recent Advances in the SISSO Method and Their Implementation in the SISSO++ Code. J Chem Phys 2023, 159 (11), 114110.



(15)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Pipeline Pilot. https://www.3ds.com/products-services/biovia/products/data-science/pipeline-pilot/.



(16)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Pipeline Pilot | Integration of Python and Jupyter Notebook. https://www.youtube.com/watch?v=1sFaA7Fj0oM, accessed on 18/09/2024.



(17)&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Yang, Z.; Dennis-Smither, B. J.; Buda, C.; Easey, A.; Jackson, F.; Price, G. A.; Sainty, N.; Tan, X.; Xu, Z.; Sunley, G. J. Aromatic Aldehydes as Tuneable and Ppm Level Potent Promoters for Zeolite Catalysed Methanol Dehydration to DME. Catal Sci Technol 2023, 13 (12), 3590–3605.







Interested in staying up to date on all the latest news of BIOVIA?




Subcribe to our newsletter!

 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Next-Generation User Assistance: Making Your Life Easier ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/next-generation-user-assistance-making-your-life-easier/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/271901</guid>
      <pubDate>Thu, 07 Nov 2024 12:56:07 GMT</pubDate>
      <description>
      <![CDATA[ Materials Management is BIOVIA’s new cloud-based solution that enables you to capture, define, and locate any proprietary or commercially sourced chemical or biological material.
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Materials Management is BIOVIA’s new cloud-based solution that enables you to capture, define, and locate any proprietary or commercially sourced chemical or biological material. Built on the powerful 3DEXPERIENCE platform, this new app provides seamless access to critical information for efficient materials management. Take a look at past blogs here and here to find out more about its capabilities and ability to capture a diverse range of entities including PROTACs and molecular glues.



Materials Management and other BIOVIA apps exist as convenient, modern web apps on the 3DEXPERIENCE platform. Interoperable and frequently updated they exploit the latest approaches and best practices available in the world of user assistance to help make usage as easy as possible.



Discover Dynamic “What’s New” Content



Gone are the days of static release notes! With Materials Management, you can now quickly see the new capabilities delivered in each release. Linked directly from the app home page, What&#8217;s New lets you very quickly get up to speed with the new features provided.







What’s New content for the Materials Management R2024x FD03 release, viewable as a set of interactive tiles.



Tiles, illustrated with an image or animated gif provide easy-to-digest highlights that explain the value of each new feature. For each one, you can click on the “Learn more” link to open a panel that provides more information about the new feature that you’re interested in, as well as a link to the detailed related topics in the Materials Management User’s Guide.







Click the “Learn More” link to open a detailed description of the selected feature.



Context Sensitive Help



Have you ever wanted to see help information tailored to a specific operation?



In Materials Management, the User Assistance panel is context-sensitive, showing you the most relevant information for the operation you’re currently working on.



Are you creating a substance and you need to generate it from the BIOVIA Basics Substances list? No problem, simply click the help icon to bring up context-sensitive guidance that explains just how to do that.







Context-sensitive help (on the right) helps you with the task in hand.



If you keep the User Assistance panel open, the context-sensitive help guidance changes as you navigate through the app, so you can always see the content you need to help you accomplish the task in hand.



Role Cockpits



BIOVIA organizes apps into roles enabling you to execute and complete key business processes, such as formulation development. The Formulation Scientist role brings together apps such as Materials Management, Scientific Notebook, Laboratory Operations, and others, so that you can undertake formulation design studies, helping to evolve and respond to changing market needs. A key element of roles like Formulation Scientist is their Role Cockpit.







The Formulation Scientist role cockpit provides scientists with access to key apps and introductory information, to enable them to be productive within minutes.



This role cockpit is designed to provide everything that a formulation scientist needs to accomplish their daily work, out-of-the-box. It provides information on what the role enables along with introductory explanations on all its apps. What’s New information is provided for each app, particularly highlighting those whose key capabilities impact the role the most.



For the Formulation Scientist role, dedicated Design, Make, and Manage Materials tabs offer dashboard based, quick access to the key apps required to enable scientists to rapidly become productive with formulation development capabilities.



Through role cockpits such as this one, scientists can understand and become proficient with formula design within minutes!



Increase Productivity Faster



Courtesy of the 3DEXPERIENCE platform, BIOVIA apps and roles embrace the latest approaches to user assistance to make it as easy as possible for you to accomplish your daily tasks!







Interested in staying up to date on all the latest news from BIOVIA?




Subscribe to our Newsletter!

 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Vratin Srivastava’s BIOVIA Internship: Iterative Protein Design Using Generative Models ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/internship-iterative-protein-design-using-generative-models/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/270030</guid>
      <pubDate>Fri, 25 Oct 2024 11:44:46 GMT</pubDate>
      <description>
      <![CDATA[ Vratin Srivastava just completed a successful summer internship on computational protein design within the Protein Modeling and Simulation team at BIOVIA….
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Vratin Srivastava just completed a successful summer internship on computational protein design within the Protein Modeling and Simulation team at BIOVIA.



During his internship, Vratin leveraged state-of-the-art generative machine learning models to create an end-to-end Python pipeline for designing novel peptide binders to protein targets. He utilized Bayesian optimization techniques and iteratively refined binder designs.&nbsp; This approach aimed to develop a pipeline that could generate peptide designs with minimal parameter input from users. The project explored whether Bayesian optimization could potentially improve the efficiency of the design process compared to methods that generate and filter large numbers of designs.




Everyone in the organization, especially Rohith Mohan and Reed Harrison who guided me through my internship and other scientists working in the modeling and simulation team, were extremely generous, extremely helpful. It was a very rewarding experience, and I&#8217;m really thankful for the opportunity to work on this project.




Curious to learn more about Vratin’s project? Watch the video below where Vratin shares details about his project and his experience working with the BIOVIA team.












 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ The rise of autonomous ships ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/industries/marine-offshore/the-rise-of-autonomous-ships/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/270433</guid>
      <pubDate>Thu, 10 Oct 2024 11:54:55 GMT</pubDate>
      <description>
      <![CDATA[ We’re entering an exciting new era in the maritime industry where ships increasingly rely on smart technology – sensors, artificial intelligence and machine learning systems – rather than crew to safely navigate the seas. 
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
In March 2023, Kongsberg’s Yara Birkeland, one of the world’s first fully autonomous electric container ships, completed its maiden voyage in Norwegian waters, entirely self-guided. It was a significant milestone in the development of autonomous shipping technology, proving that a vessel could safely and efficiently navigate itself in real-world conditions.



Today, under human supervision, the vessel transports containers from a production facility to a deep-sea container harbor. It autonomously leaves the quay, sails across the water and docks at the other side. A small onboard crew supervise and monitor the ship, but this is mainly to satisfy regulatory requirements, which currently mandate a certain level of human presence onboard.



Equipped with sensors, artificial intelligence (AI) and machine learning systems, this vessel is still only one of a few autonomous ships operating in open waters, yet many more are expected to enter into commercial operation in the near future. It’s hoped that rapid technological innovation combined with evolving regulatory frameworks will pave the way for safer, more sustainable and efficient shipping.



Underwater autonomous drone







What is autonomous shipping?



Moving from manual to autonomous vessel operations involves developing all the capabilities needed to operate with little to no human intervention. Today, ships are increasingly equipped with advanced technologies like integrated automation systems and sensors, which allow them to navigate, make decisions, and perform tasks independently.



The International Maritime Organization (IMO) sets out four stages of autonomy for maritime autonomous surface ships (MASS), ranging from vessels with some automated processes through to a fully autonomous vessel that operates without human intervention: &nbsp;




Degree one: Ship possesses automated processes and decision support but has seafarers onboard for backup operation and control.



Degree two: Ship is controlled entirely remotely yet seafarers remain onboard.



Degree three: Ship is remotely controlled without seafarers onboard.



Degree four: Fully autonomous ship makes decisions and determines actions by itself. &nbsp;




Today, most ships in testing are at level two; they have the capabilities to operate autonomously but still have crew onboard. Until regulations catch up, it’s likely that the industry will remain at this stage for some time. Technical and infrastructure issues also remain, such as ensuring a continuous connection with the vessel at sea. &nbsp;











What are the benefits of autonomous shipping?



It’s widely recognized that autonomous technology will bring substantial benefits to ship owners, particularly in terms of improving safety, cost efficiency and meeting the UN’s sustainable development goals.




Safety: One of biggest drivers of increased autonomy is to improve safety at sea. A large proportion of accidents at sea are caused by human error – insurance firm Allianz estimates it accounts for 75% to 96% of accidents. Autonomous technology promises enhanced situational awareness, navigation and collision avoidance.



Efficiency: Automated operations can help to optimize routes, travel time and fuel consumption. Taking human error and fatigue out of the equation reduces downtime. Autonomous cargo handling systems can also streamline offloading/unloading and reduce damage to cargo.



Reduced costs: Cost efficiencies come from both decreased capital and operational expenditure; for example, lowering fuel consumption as a result of better voyage planning and execution. Companies can also expect to save costs by reducing crew and support workers, most notably on smaller ships engaged in near-coastal operations, including small island ferries, tugboats, barges and supply and service vessels.



Sustainability: Most autonomous ships in development and operation today run on electricity and alternative fuels, helping to reduce emissions and pollution. Optimized routes improve fuel efficiency. Fewer accidents and collisions will also lead to less environmental damage such as oil spills.








What are the disadvantages of an autonomous ship?



IMO has raised a range of issues linked to autonomous shipping, including safety, security, liability and compensation for damage, interactions with ports, pilotage, responses to incidents and protection of the marine environment.



Key industry stakeholders will need to balance the benefits of autonomous shipping technologies against potential challenges such as:




Technological limitations and risks: Autonomous systems will need to operate in sometimes extreme conditions, including adverse weather and areas of high traffic. Autonomous ships are also vulnerable to cyberattacks, risking safety and continuous operation.





Regulatory hurdles: Developing comprehensive regulations for autonomous shipping is proving complex and time consuming. Regulatory bodies will need to develop frameworks that prioritize safety and keep pace with fast evolving technology developments.





Ethical issues and public acceptance: Autonomous technologies raise questions about risks to jobs, liability and decision-making in critical situations. Public concerns around safety and reliability also need to be addressed.





Cost: Rolling out autonomous shipping systems requires high upfront investment associated with technology development, research, infrastructure and regulatory compliance.








Which types of vessels are better suited to being autonomous?



Regulatory restrictions and technological limitations mean that the first vessels best suited to going autonomous will be those operating in controlled environments over short distances, such as from one port to another along coastal waters and inland waterways. That would include everything from tugboats and commuter ferries to service vessels such as for windfarm maintenance and inspection. Here, there is a stronger business case for the technology.



Deep sea activities like cargo shipping will come later as the technology matures and associated development costs come down.



How must regulations support the future of autonomous shipping?



Developing entirely new regulations for emerging technologies takes time. Classification societies and industry regulators like IMO have made clear their intentions to support the development of autonomous shipping and have committed to working together to define industry-wide rules. The current plan is for a non-mandatory MASS code to come into effect in 2025, followed by a mandatory MASS code in 2028.



In the meantime, IMO has several high-priority issues to address including:




Terminology and definitions: Establish clear definitions for MASS as well as the responsibility of certain personnel such as master and crew.





Operational requirements: Determine the role of remote control centers and operators.





Safety treaty gaps: Review where there may be a need for manual operations as well as implications for search and rescue, watchkeeping, firefighting, security and maintenance.




As shipbuilders contend with how to integrate autonomous technologies into future ship designs, they will need to master increasingly complex systems engineering challenges. Dassault Systèmes offers powerful solutions to efficiently model, simulate, and validate concept designs, integrating both technical and business data. Model-based systems engineering and system of systems engineering solutions help to achieve complex multi-disciplinary systems modeling and, through problems analysis, rapidly solve design challenges early on.



With the 3DEXPERIENCE platform, mechanical, electrical, fluidics and software disciplines are implemented in context of the system requirements and wider architecture, providing a holistic view of the overall ship’s system design. Teams can collaborate from a unified information source to make informed tradeoffs and develop the best concept that also complies with industry regulations.



While autonomous technologies continue to advance, work remains before unmanned ocean-going vessels become a common sight. Public acceptance and awareness of what autonomous technologies can and cannot do must also grow. Likewise, the technology must prove itself and come down in cost for it to become truly viable. Be that said, things are moving at a fast pace, and we should expect to see more remotely operated vessels entering operation in the next couple of years.







Discover more:



Maritime fuels of the future



From ADAS to Autonomous Driving



Could nuclear energy decarbonize cargo shipping?
 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Biotherapeutics: What Do We Make Next? ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/biotherapeutics-what-do-we-make-next/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/270366</guid>
      <pubDate>Wed, 09 Oct 2024 18:39:22 GMT</pubDate>
      <description>
      <![CDATA[ For years, computational methods for small molecule drug design have offered numerous algorithms and methodologies to help generate new ideas and guide the iterative process of lead design and…
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
The question of what should we make next has challenged the world of drug discovery for decades. For years, computational methods for small molecule drug design have offered numerous algorithms and methodologies to help generate new ideas and guide the iterative process of lead design and optimization. For a particular drug target, these methods help to identify high-quality candidates that may eventually advance to clinical development with less experiments and time in the lab. From the early days of combinatorial chemistry and bioisosteric replacement to ligand-, fragment- and structure-based design, there have been many tools, leveraging numerous algorithms that suit your project constraints and design criteria.&nbsp; More recently, AI and machine learning algorithms have been popular in allowing researchers to rapidly explore more ideas in the chemical space and propose novel structures that a medicinal chemist may not have considered trying out when looking for new drugs.



Until recently, the computational design tools for biotherapeutics seemed to require more expertise, and to be more sparse and application-specific compared to the tools that exist for small molecule therapeutics. Of course, there are computational design algorithms available such as homology modeling, protein-protein docking and combinatorial scanning mutagenesis for general protein modeling and binder design, which are used in biotherapeutics lead discovery and optimization. For designing certain types of biological therapies, such as monoclonal antibodies, there are methods such as affinity maturation, humanization and immunogenicity prediction algorithms. However, to help answer directly what variation of our biotherapeutic we should make and test next, two recent AI methods, RFDiffusion and ProteinMPNN, have totally changed the nature of biotherapeutics discovery. These tools have the potential change the way we design biotherapeutics by helping to identify novel candidates that the computational and molecular biologists may not have considered.



Generating Proteins with AI: RFDiffusion and ProteinMPNN



RFDiffusion is a cutting-edge generative AI algorithm that can &#8220;diffuse&#8221; a collection of amino acids into a protein structure. The diffusion process starts with a random, noisy collection of atoms and, through a series of controlled refinements the algorithm makes adjustments to the structure to reduce the noise and move closer to a biologically realistic and functional protein structure. One common analogy for the diffusion process is developing a photo from a blurry image; iterative processing steps can take an initial grainy image and refine the detail and clarity to produce a final clear picture.



RFDiffusion can be utilized for a number of different biotherapeutic design challenges, such as engineering a biologic that can bind to a viral protein to neutralize the virus. With antibody structures or other protein-protein systems, RFDiffusion can be used to design new protein scaffolds that may improve binding affinities or enhance the stability of the binding partners. RFDiffusion can be also used to generate enzyme therapeutics that may break down a specific substrate to treat metabolic disorders. Beyond biotherapeutics, RFDiffusion has potential to help design proteins for industrial and biotechnological applications such as making enzymes that catalyze specific chemical reactions or proteins that suit very specific conditions including low or high temperature, pH, etc.



ProteinMPNN is a state-of-the-art neural network that can predict one or more probable protein sequences given a protein structure. This algorithm has been published with success in one of the most critical aspects of protein sequence design – generating sequences that fold into a stable protein/peptide with propensity to crystallize, facilitating the structure determination of these proteins. ProteinMPNN can be used in conjunction with RFDiffusion to generate new protein designs such as new enzymes or antibodies that can be further evaluated for desired properties such as stability, activity, affinity, and specificity. One of the strengths of ProteinMPNN is its ability to generate multiple sequence variants. This ability is invaluable as different variants provide more options to test and identify candidates with the best performance in terms of efficacy, safety, and manufacturability. Just as significantly, these variants also provide alternative leads when candidates encounter unforeseen issues in protein optimization, during protein expression, or ADMET challenges such as solubility and immunogenicity.



Together, RFDiffusion and ProteinMPNN significantly expand the biological space that can be explored in silico before biologists need to commit to expensive and time-consuming physical experimentation.  They have the potential to open up exciting avenues for more intelligent, model- and data-driven workflows driving innovation in biotherapeutic design.



Generating Proteins with RFDiffusion and ProteinMPNN in Discovery Studio Simulation



In BIOVIA Discovery Studio Simulation, a new Generate Protein Scaffolds protocol now provides easy access to RFDiffusion workflows, the first of which is motif scaffolding. Users can start with a specific part of an existing protein (the motif) and design a complete new protein scaffold that incorporates this motif. This approach allows precise control over the functional regions of the protein, as well as control over the protein scaffold design, via different model weights that suit particular proteins and complexes.



Figure 1- Discovery Studio Simulation users now have access to motif scaffolding with RFDiffusion.



A second new protocol, Generate Protein Sequences, allows users access to not only ProteinMPNN, where they can easily define sequence residues for design, but also to LigandMPNN and SolubleMPNN models. LigandMPNN is an extension to ProteinMPNN that is able to consider protein, small-molecule, nucleic acid, and metal ion ligands as additional context for designing sequences, with the potential to improve the chemical properties of the designed sequences. SolubleMPNN could be a better model to use when protein solubility is part of your design criteria. Users can determine the degree of sequence diversity and confidence desired, as part of the generative design, and have the ability to control the bias of particular amino acids.



Figure 2- Discovery Studio Simulation users can now generate new sequences using ProteinMPNN models and use AlphaFold/OpenFold to generate their 3D structures for further applications.&nbsp;







These two significant new enhancements are exciting additions to the biotherapeutics and protein design tools in Discovery Studio Simulation in the 3DEXPERIENCE® Cloud, which already includes AlphaFold and OpenFold AI structure prediction. They expand the ever-growing arsenal of powerful AI tools for molecular modelers and biologists to help answer the question of “what to make and test next” and accelerate the rational design of biologics. In combination with the existing physics-based methods in Discovery Studio Simulation, users can rapidly explore many more possibilities in silico before arriving at the final handful of candidates that are ready to become a successful commercial biotherapeutic or a biological to be used in agriculture, food and beverage, or environmental industries.



Nobel Prizes in Chemistry and Physics



This year’s Nobel Prizes in Chemistry and Physics celebrate how AI is pushing the boundaries of scientific research. John J. Hopfield and Geoffrey E. Hinton were awarded the Nobel Prize in Physics for their foundational discoveries in machine learning with artificial neural networks, while David Baker, Demis Hassabis, and John Jumper received the Nobel Prize in Chemistry for breakthroughs in computational protein design and protein structure prediction.



At BIOVIA, we are proud to be part of this AI revolution. By integrating AlphaFold2, OpenFold, RFDiffusion, and the ProteinMPNN family of models into our platform, we empower researchers with cutting-edge tools for protein structure prediction and protein design.



Watch the video to learn more how Discovery Studio Simulation now helps users generate novel biologics with RFDiffusion and LigandMPNN models.
















 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Precision Polymer Modeling: Leveraging Materials Studio and Scripting Innovations ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/precision-polymer-modeling-leveraging-materials-studio-and-scripting-innovations/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/270066</guid>
      <pubDate>Wed, 09 Oct 2024 15:04:39 GMT</pubDate>
      <description>
      <![CDATA[ Materials Studio offers a user-friendly yet powerful platform for modeling a huge range of systems, with this report focusing specifically on polymers and polymer networks…
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Materials Studio offers a user-friendly yet powerful platform for modeling a huge range of systems, with this report focusing specifically on polymers and polymer networks. In conjunction with Pipeline Pilot, various methods are available to model such complex materials ranging from the widely reported cross-linked polyethylene (XLPE), to custom networks with unique reaction mechanisms.&nbsp; United atom forcefields are developed to efficiently model such systems involving large numbers of particles, providing a balance between computational efficiency and accuracy.



The Importance of Forcefields in Molecular Simulation



A critical element of molecular simulation is the forcefield.&nbsp; Forcefields govern the behaviour of particles in a system through both bonded and non-bonded interactions, covering contributions including bond stretching, angle bending, dihedral torsions, etc.&nbsp;In polymer models, the forcefield is essential for determining how polymer chains fold, entangle with one another and respond to external stimuli. By accurately modelling these atomic-level interactions, forcefields enable the prediction of macroscopic properties such as density, radius of gyration, and glass transition temperature. Thus, an effective forcefield is essential for both the accuracy of simulations and the reliability of the predicted properties.



There are a number of forcefields available depending on the type of system being modelled.&nbsp; Some forcefields are optimized for specific materials while others, like COMPASSIII [1], are designed for a broader range of materials. The field of forcefield development is continually developing as exemplified by the recent emergence of MACE [2]; a machine learning software, which is used, among other things, to generate forcefields.&nbsp; Of particular significance in polymer modelling are OPLS and OPLS-UA.&nbsp; OPLS, Optimised Potentials for Liquid Simulations [3], was specifically designed to simulate liquids and later expanded to cover a wide range of organic molecules, biomolecules, and polymers.&nbsp; OPLS-UA is a modified version of this forcefield for a united atom approach.&nbsp; Given that molecular simulations can be computationally expensive, especially for systems with large numbers of particles, it is often desirable to simplify these systems by reducing particle counts while simultaneously minimizing the loss of accuracy. This simplification is achieved through united atom or coarse-graining methods These approaches also provide a further benefit of longer time steps, allowing users to conduct simulations with more steps within a set timeframe.



Custom Forcefield Development in Materials Studio



In Materials Studio, users create custom forcefields by inputting parameters in various functional forms, gaining full control. Through literature review and modification of OPLS atomistic parameters, followed by iterative testing and refinement, an OPLS-UA custom forcefield was developed [4]. Torsion parameters, for instance, were validated using conformational analysis on single molecules. The integration of Perl scripting within Materials Studio provided precise control, and a simple script was created to extract torsion energy as a function of torsion angle. These results were then compared to literature data or DMol³calculations. Once refined, the forcefield was tested on six polymer systems, with an Amorphous Cell constructed for each system and equilibrated through molecular dynamics simulations. Physical properties of the equilibrated systems were extracted and compared to literature values. One key test was the radius of gyration as a function of the degree of polymerization, which is well documented for many polymers. To automate the process of generating cells, equilibration, and calculating the radius of gyration for systems with incrementally larger degrees of polymerization, custom Pipeline Pilot protocols were developed [5]. This integration significantly reduced time and minimized human error. It allowed for the radius of gyration data to be compared with theoretical values and ultimately validate the accuracy of the developed forcefield, as exemplified in Figure 1.







Alongside the creation and validation of the forcefield, a simple method was developed to convert a polymer from full atomistic to a united atom representation.&nbsp; This was achieved by removing hydrogen atoms, adjusting the mass numbers of the corresponding beads, and assigning the appropriate forcefield type to each bead. A custom Perl script was created for this process, using a Study Table to perform a pattern search through the included fragments.&nbsp; Each fragment corresponds to a different atom type, such as sp³ CH₂, sp² CH, or aromatic CH, along with their associated forcefield types.&nbsp; As a result, only the name of the input 3D Atomistic Structure file needs to be inserted into the script for the polymer to be automatically converted to united atom and typed.



With a method in place to convert and type polymers, along with the validated OPLS-UA forcefield, polymer and polymer network modelling using this forcefield could be further explored. Depending on the system being constructed, various reactions need to be modelled—most notably addition, cycloaddition, and condensation reactions.&nbsp; Materials Studio offers several methods for this purpose, with the choice depending on the central reaction. Three main methods were employed to model these systems: a customizable networking Perl script, a Pipeline Pilot protocol, and the Reaction Finder tool.&nbsp; Examples of each are discussed below.



Cross-linked polyethylene (XLPE) is a widely studied polymer network, used in numerous applications ranging from electrical insulation in wires to components in fuel gaskets, owing to its high chemical and heat resistance.&nbsp; Users can model this network in Materials Studio using the Pipeline Pilot connector with the modified Network Protocol, as demonstrated in Figure 2. The process begins by creating an Amorphous Cell containing both ethane molecules and polyethylene chains, with reactive atoms defined. These reactive atoms dictate the reactions that build the network, allowing the polymer chains to grow linearly, branch out, or cross-link with other chains. The probability of each reaction is specified, providing additional control over the end system.&nbsp; By altering the probability of the defined reactions, cross-linking density is modified and the effect on physical properties after equilibration can be investigated, as exemplified in Figure 3.







Figure 2: Fragment from XLPE displaying cross-links formed between chains.  The atoms are coloured by forcefield type.



Figure 3: Effect of the extent of cross-linking in equilibrated XLPE models on the density. 



Diels-Alder reactions are prominent cycloaddition reactions between a diene and an alkene, and can be used to join monomers in more complex polymer systems. The Reaction Finder[6] tool is employed to precisely define these reactions by drawing both the reactants and products, mapping the atoms between them, and identifying close contacts between the reactive atoms. This is exemplified in Figure 4.&nbsp; This approach offers significant control, making it particularly useful for modelling specific structures with atypical reactive sites.



Figure 4: Reactants (left) with close contacts defined for a Diels-Alder reaction to form products (right).



To model condensation reactions, a customizable networking Perl script[7]&nbsp; was adapted and used effectively. Reactive atoms are defined in both the monomer and curing agent, which are then organised into an Amorphous Cell. The script specifies these reactive atoms along with the input structure document. An additional subroutine within the script is used to define specific bond-breaking events during the reaction, producing the desired product as shown in Figure 5. This versatile script supports the cross-linking of various monomers and curing agents and can be tailored to meet different system requirements.



Figure 5: Reactants (left) with reactive atoms R1 and R2 defined to form the product (right).



Conclusion: The Power of Materials Studio for Modeling Polymer Networks using United Atom Forcefields



In conclusion, Materials Studio proves to be an exceptional platform for modelling complex polymer systems and networks, offering a wide range of tools and customization options. Its integration with Pipeline Pilot, the ability to develop and refine custom forcefields, and the flexibility provided through Perl scripting enable precise control over model building and simulations.&nbsp;&nbsp;The platform&#8217;s versatility and the high degree of control it offers open up exciting possibilities for future research, allowing users to push the boundaries of polymer design, optimization, and property prediction with great accuracy and efficiency.







[1] Akkermans, R. L. C., Spenley, N. A. and Robertson, S. H. (2020) ‘COMPASS III: automated fitting workflows and extension to ionic liquids’,&nbsp;Molecular Simulation, 47(7), pp. 540–551, doi: 10.1080/08927022.2020.1808215.



[2] https://mace-docs.readthedocs.io/en/latest/ (accessed 09/2024).



[3] William L. Jorgensen, David S. Maxwell, and Julian Tirado-Rives (1996), ‘Development and Testing of the OPLS All-Atom Force Field on Conformational Energetics and Properties of Organic Liquids’, Journal of the American Chemical Society&nbsp;,&nbsp;118&nbsp;(45), pp. 11225-11236, doi: 10.1021/ja9621760.



[4] Available in the BIOVIA Materials Studio community.



[5] Available in the BIOVIA Materials Studio community.



[6] J.W. Abbott and F. Hanke, &#8216;Kinetically Corrected Monte Carlo-Molecular Dynamics Simulations of Solid Electrolyte Interphase Growth&#8217;,&nbsp;J Chem Theor Comput,&nbsp;18, 925 (2022).



[7] Developed by Jason DeJoannis, Stephen Todd &amp; James Wescott (available in the BIOVIA Materials Studio community).












 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ Miko Stulajter’s BIOVIA Internship: Prototyping and New Features in BIOVIA Molecular Design ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/biovia/miko-stulajters-biovia-internship-prototyping-and-new-features-in-biovia-molecular-design/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/269774</guid>
      <pubDate>Tue, 01 Oct 2024 08:28:00 GMT</pubDate>
      <description>
      <![CDATA[ Miko Stulajter completed a second three-month summer 3DEXPERIENCE® scientific visualization internship with the BIOVIA R&D software engineering team. After a rewarding first internship experience…
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Miko Stulajter completed a second three-month summer 3DEXPERIENCE® scientific visualization internship with the BIOVIA R&amp;D software engineering team. After a rewarding first internship experience, Miko was excited to return for another summer, attracted by the meaningful work he had accomplished through his first internship.



During this internship, Miko focused on implementing and prototyping new features for visualizing molecular trajectories in BIOVIA Molecular Design. Miko developed a new trajectory animation dialog UI, significantly improving user control over trajectory playback. He also prototyped a streaming server to facilitate efficient testing and development for reading remotely hosted trajectory files.&nbsp; Additionally, Miko explored utilizing D3.js to create interactive trajectory visualizations, such as a timeseries of trajectory properties and a potential energy surface visualization of trajectory frames, enhancing the overall trajectory analysis experience in BIOVIA Molecular Design.



Beyond his technical contributions, Miko’s quick learning and adaptability allowed him to rapidly develop and prototype these new features, making him an invaluable part of our intern program this year.




“I really enjoyed my first internship with BIOVA and was excited to return for a second one. It was fulfilling to work on new projects and see some of the work I did last summer now in production. Coming back to BIOVIA felt like I never left.”




Curious to learn more about Miko’s project and his experience working with the BIOVIA team? Watch the video below, where Miko shares his work and the insights he gained.








 ]]>
      </content:encoded>
      </item>
<item>
      <title>
      <![CDATA[ What is NLP? (Natural Language Processing) ]]>
      </title>
      <link>https://blog--3ds--com.apsulis.fr/brands/netvibes/what-is-nlp-natural-language-processing/</link>
      <guid>https://blog--3ds--com.apsulis.fr/guid/269333</guid>
      <pubDate>Thu, 19 Sep 2024 05:15:00 GMT</pubDate>
      <description>
      <![CDATA[ Explore the benefits and challenges of NLP and how it is revolutionizing industry. 
 ]]>
      </description>
      <content:encoded>
      <![CDATA[ 
Ever feel like a chatbot just gets you? It’s not magic. It’s … science.



Natural language processing is a powerhouse transforming relationships between humans and technology. It helps bots understand our questions, sifts through massive amounts of unstructured data and performs advanced communicative tasks. NLP is trained on advanced algorithms to understand, manipulate and generate the human language.



Is NLP a new concept?



In recent years we have become more accustomed to AI and the pairing of language and technology with the influx of large language models like ChatGPT, yet the science and process behind them, including NLP, has been around for decades.



NLP uses algorithms to analyze textual relationships through language analysis and comprehension, while LLMs use deep learning to mimic human language and generate text. While tools like ChatGPT are relatively new, NLP has been around since the mid-20th century. It initially focused on rule-based systems in the 1950s and evolved into statistical models in the 1990s.



Natural Language Processing, defined



Natural language processing is a subfield of computer science machine learning. It enables computers to understand and communicate with human language. NLP evolved from computational linguistics which utilizes computer science to understand the very principles of language. NLP works with computers and other devices to recognize, understand and generate text &amp; speech by combining computer-based modeling of human language with statistical modeling, machine learning, and deep learning.



The advancement of NLP is enabling its integration into diverse fields such as healthcare, finance manufacturing and customer service, enhancing human-computer interactions and shaping the future of AI-driven communication technology.



“It’s a little bit like a human. It goes through documents highlighting the words and forms of expression that are important as defined by the classification plan, allowing us to quantify the various concepts,” said Kelly Stone, an NLP expert for Dassault Systèmes’ Information Intelligence brand, NETVIBES.  



Categories of NLP



NLP can be divided into three main categories regarding its various tasks and applications. When deciding what NLP works best for your business consider what task you aim to achieve. Below are three main subcategories of NLP:




Rules Based NLP: Rules-based NLP were the earliest NLP applications that answered simple if-then decision trees requiring pre-programmed rules. They were only able to provide answers in response to specific prompts.



Statistical NLP: Statistical NLP extracts, classifies, and labels elements of text and voice data and then assign a statistical likelihood to each possible meaning of those elements. This form of NLP introduced the technique of mapping language elements such as words and grammatical rules.



Deep-learning NLP: Deep-learning NLP is the dominant mode of NLP most users interact with which uses huge volumes of raw unstructured data to become more accurate. Deep-learning NLP is a further evolution of this statistical NLP.




How does NLP work?



NLP works like a digital linguist, deciphering the intricate patterns and meanings embedded in human language. It starts by breaking down sentences into smaller components, like words and phrases, and then dives deeper to understand grammar, semantics, and context.



Through machine learning algorithms and vast datasets, NLP learns to recognize pattern usage, enabling it to perform tasks such as sentiment analysis, language translation, and speech recognition. By constantly evolving and learning from new data, NLP works to adapt to nuances and changes in language over time.



NETVIBES is currently using NLP to help companies across industries overcome a number of unstructured data issues. For example, to review customer satisfaction surveys regarding a hotel for a client which involve massive amounts of unstructured data. Categories are created such as cleanliness, safety, and comfort. The model can then identify concepts in the customer reviews ranking them as positive, negative or neutral within many subcategories. The ranking of each category is then produced as a percentage of negative and positive reviews and an overall customer satisfaction percentage is derived from these subcategories.



How is NLP transforming business?



NLP has become a part of most of our everyday lives, working to power search engine results such as Google, providing customer service chatbots, and driving with voice-operated GPS systems. NLP has had a growing role in enterprise solutions, streamlining and automating business operations, increasing employee productivity, and critical business processes.



NLP is continuously being applied to diverse fields like retailing for customer service, chatbots and medicine, interpreting and summarizing electronic health records. Conversational agents such as Amazon’s Alexa also utilize NLP to listen to users and find answers.



In the healthcare field, NLP accelerates the process of reviewing and extracting relevant data from research papers, aiding in the discovery of new treatments and understanding of diseases. Chatbots and virtual assistants powered by NLP provide patients with information, schedule appointments, and offer preliminary health advice, enhancing patient engagement and accessibility.



With NLP baked into its solutions, NETVIBES is helping companies analyze large amounts of data and discover insights, monitor employee and customer experiences, and streamline business processes for previously tedious tasks.



“About 90% of a company’s data is unstructured, making it very difficult to create value from,” said Stone. “NLP can analyze any unstructured data, ranging from customer experience data such as surveys, email complaints, and help companies to quantify what is driving satisfaction and make action plans to improve the customer experience. It can analyze change requests and quality reports to help companies optimize their internal processes and improve quality”.



How can your business use NLP?



NLP makes tedious tasks easier by taking massive amounts of unstructured data and make sense of it. But NLP does not stop there, here are additional values the technology holds according to insights from DeepLearning.AI.




Linguistic tasks: This involves identifying if and when two words refer to the same entity.



Part of speech tagging: NLP determines which part of speech a word or piece of text is based on its use and context.



Word sense disambiguation: This selects a word meaning for a word with many possible meanings.



Named entity recognition: NLP identifies words or phrases as useful entities when scanning large datasets.



Spam detection: Large email services like Gmail use prevalent binary classifications to determine whether emails are spam or not. This allows for a better user experience removing unwanted emails from our inboxes.



Online grammar checkers: Grammar checkers like Grammarly use such systems to provide better writing experiences offering insights for grammatical corrections for writers to incorporate. These platforms also have helped teachers grade students&#8217; essays in the classroom.




Five major benefits of NLP



Once properly trained, NLP models can work rapidly and effectively and take on tasks for workers focusing their attention on other areas.




Faster business discovery: NLP uncovers hidden relationships between different pieces of content. Through text data retrieval, deeper insights and analysis enable better informed business decisions.



Cheaper data processing: NLP automates data gathering and processes information with less manual effort, decreasing human labor costs. When businesses have a massive volume of unstructured text data to sift through, this information can be easily categorized and understood.



Automation of tasks: NLP automates routine tasks such as customer support queries, content generation, and data extraction. This increases efficiency in business and production streamlining previously tedious tasks.



Language translation: This technology bridges communication gaps across languages facilitating global interactions and commerce. NLP is breaking down the barriers in understanding across businesses.



Improved accessibility: NLP enables accessibility features like speech-to-text and text-to-speech for people with disabilities. It further improves users experience through customization of user preference based on language and behaviors, enhancing engagement.








Why is NLP difficult?



NLP models remain imperfect and likely will never reach any level of perfection, similar to how humans continue to learn language their entire lives.




Biased training: If exposed to bias data in training, NLP similar to other AI functions will result in skewed answers. One way to overcome this is to train NLP functions on more diverse datasets. However, training datasets that are often scraped from the web are prone to bias.



Misinterpretation: In AI there is also a risk of misinterpretation due to the lack of clear quality input involving mumbles, slang, or other mispronunciations. The input to the tool is critical to ensure misinterpretations are few and far between.



New vocabulary: With new words being invented or imported NLP can only make its best guess or admit it is unsure. These datasets need to constantly be updated and trained to ensure that new conventions and ways of speaking are incorporated into the NLP tool.



Ambiguity in language: When words and phrases have multiple meanings depending on the text, this ambiguity can make it challenging for NLP systems to accurately interpret and generate human-like responses.




The main difficulty isn’t necessarily with technology, but rather the complexity of human language, explains Stone. “We don’t always realize how complex language is until we’re trying to learn a second language or misinterpreting the meaning of a text due to missing context,” she said. “



Addressing these challenges of NLP requires advancements in machine learning, natural language understanding, and the integration of broader contextual information to enhance the capabilities of NLP systems.



What is the future of NLP?



NLP is paving the way for smarter and more personalized interactions, from healthcare to customer service to entertainment. This is a new era of seamless communication and collaboration in the digital age.



Here at Dassault Systèmes, NLP understands human language at a deeper level unlocking data previously hidden in unstructured text. NLP first gained traction here through the 2020 acquisition of Proxem, a France-based specialist in AI-powered semantic processing software. NLP has since expanded into the 3DEXPERIENCE platform working alongside NETVIBES Information Intelligence applications. This platform delivers a combination of rule-based natural language understanding, natural language processing, and machine learning technologies to see and understand the bigger picture.



NETVIBES uses NLP every day to support clients in making meaning from large amounts of data. They have also introduced their version of ChatGPT that will work internally trained on their specific datasets providing more accurate information to clients and businesses. Indeed, thanks to this kind of technology Dassault Systèmes will be able to offer conversational assistants for augmented employee based on a Retrieval-Augmented Generation (RAG) that takes into account all the knowledge and instanced information from the different application of the 3DEXPERIENCE platform.



The future of NLP holds immense promise, driven by advancements in machine learning and AI. We can expect increasingly sophisticated models that understand and generate human-like text and comprehend context, tone, and nuance with greater precision. As NLP continues to evolve, ethical considerations around data privacy, bias, and the responsible use of AI will become increasingly important, shaping how these technologies are integrated into society and our business.
 ]]>
      </content:encoded>
      </item>
    </channel>
   </rss>