An article just appeared which summarizes (and includes case examples for) some important lessons for the application of computational methods in drug design. I write about it here because it includes important messages also for the materials modelling and design field. There are differences of course since materials innovation is not always about new materials design. Nevertheless some of the key points are still valid and at least worth considering. I like the ‘principle of parsimony’, and also the conclusion about the importance of good software design. Much needed in the materials field as well.
Here are some key quotes and extracts from the paper .
The value of qualitative statements. Frequently, a single new idea or a pointer in a new direction is sufficient guidance for a project team. Most project impact comes from qualitative work, from sharing an insight or a hypothesis rather than a calculated number or a priority order. The importance of this observation cannot be overrated in a field that has invested enormously in quantitative prediction methods. We believe that quantitative prediction alone is a misleading mission statement for molecular design. …
Shaping chemical space. At any given point during a project, a team’s focus is either on expanding chemical space or on narrowing it down, for different aspects of problem solving and optimization. Broadening chemical space requires methods that create new ideas within a set of constraints. ….Narrowing down chemical space can be a simple filtering process or can be based on a specific hypothesis. Within a given project context, it is important to understand whether it is required to broaden or narrow down chemical space and to choose tools and approaches accordingly. As projects progress towards candidate selection, the “amplitudes” of narrowing and broadening space typically become smaller, but the concept stays the same.
The principle of parsimony. Molecular design is a conceptual process and therefore always at risk of losing touch with reality. The scientific questions should lead to the method, and not vice versa. To achieve this, it is a helpful guiding principle to keep things as simple as possible. Choosing the simplest possible explanation and the simplest possible computational protocol leads to agility and to a better focus on the key questions at hand. …
Annotation is half the battle. … Contextual information can add value almost anywhere. A good deal of frontloading work—computational, organizational—is often required to bring data into a useful shape. Proper frontloading work can turn sophisticated queries into simple lookup processes or visualization steps. There is a significant growth potential in this area.
Staying close to experiment. One way of keeping things as simple as possible is to preferentially utilize experimental data that may support a project, wherever this is meaningful. … Rational drug design has a lot to do with clever recycling. If consistently applied, these guidelines have significant implications for the current practice of molecular design.
Let us look at some of the more problematic aspects as well. Many computational methods introduce additional parameters and thus potential sources of error that make the predictive value harder to extract. …..
What is special about molecular design is the need to build solid hypotheses and to simultaneously foster creative thinking in medicinal chemistry. If we accept this, our focus may shift from the many semi-quantitative prediction tools that we have to methods supporting this creative process. Further improvements in computational methods may then have less to do with science than with good software engineering and interface design. The tools are a just means to an end. Good science is what happens when they are appropriately employed.
 A Real-World Perspective on Molecular Design. Bernd Kuhn, Wolfgang Guba, Jérôme Hert, David W. Banner, Caterina Bissantz, Simona Maria Ceccarelli, Wolfgang Haap, Matthias Körner, Andreas Kuglstatter, Christian D. Lerner, Patrizio Mattei, Werner Neidhart, Emmanuel Pinard, Markus G. Rudolph, Tanja Schulz-Gasch, Thomas J. Woltering, and Martin Stahl
J. Med. Chem., DOI: 10.1021/acs.jmedchem.5b01875 • Publication Date (Web): 15 Feb 2016
Boeing is moving ahead with integrating chemistry and materials modelling into the product life cycle
The EMMC Roadmap for Materials Modelling is calling for a number of actions to increase the application and impact of materials modelling in industry. In its Objectives and Vision it states that “the ultimate goal is that materials modelling and simulation will become an integral part of product life cycle management in European industry” and that “in recent years, materials modelling of nano-scale phenomena, especially that based on discrete models (electronic/atomistic/mesoscopic) has developed rapidly. However, this has not yet led to the integration of these models as part of the industrial design tool chain of materials and products.” While true in general, it turns out that world leading organisations are already implementing such integration.
An example of a company that seems to be well ahead of the curve is Boeing. In fact, this blog was triggered by a panel discussion that I participated in at the Predictive Materials Modelling workshop in Cambridge early December 2015. The panel on Aerospace applications was led by Airbus who presented their elaborate work on virtual testing of aircraft frames. While the work is clearly very challenging in terms of the computational resolution in FEM models required and the issues in identifying ‘hotspots’ etc., at this stage of development the materials are already well defined and no variation in material parameters are allowed or considered any more. So what about the actual materials development and its integration into this process as outlined in the vision above?
Certainly Airbus competitor Boeing has been very active in the materials modelling field down to the chemistry level for some time, see for example Reference [[i],[ii]]. Rather than relying on the traditional supply chain dynamics, Boeing has become involved in chemistry based research in-silico, thereby taking a pro-active role in shaping its own future across all disciplines.
Three recent Boeing patents actually demonstrate the significance of the corner-stones of exploitation of materials modelling at the industrial level which are also highlighted in the EMMC roadmap (a) materials modelling has been developed to a point that it can make an impact on real industrial problems, (b) multi-scale modelling workflows are key to realising impact, (c) it is important for business efficiency and effectiveness to integrate information gained down to the chemistry level into wider information management and business decision support systems.
Testimony to (a) is Patent US20150080494 (Filing date: 4 Feb 2014) “Fiber-reinforced resin composites and methods of making the same”. It deals with the efficiency of load transfer between the fiber and the surrounding matrix at the micro-scale level, which may directly affect the overall mechanical performance of the composite at the continuum level. “The region of the matrix that may be substantially affected by the presence of fibers, sometimes referred to as the “interphase” region, is the interfacial area of the matrix directly surrounding the fiber. In composites, this interphase region may experience high shear strain due to the mismatch in elastic stiffness between the fibers and the surrounding matrix. Widely-used conventional bulk resins may not provide desirable distortional capabilities.” The patent claims superior performance of resins developed with the help of atomistic materials modelling. This performance improvement could translate into substantial efficiency in load bearing and associated lower weight of the aircraft frame.
Testimony to (b) is Patent US 08862437 (Application date: 30 Mar 2010) on “Multi-scale modeling of composite structures”. The following patent abstract is a bit hard to read but basically seems to claim that there is a controlled, deterministic relationship between composite performance and materials/chemical structure at various levels, as calculated by modelling: “A method, apparatus, and computer program product are present for creating a composite structure. A number of characteristics for a number of components for the composite structure is obtained from a simulation of the composite structure using a model of the composite structure. A number of changes in the number of characteristics needed to meet a desired level of performance for the number of characteristics is ascertained. A number of attributes for a number of composite materials used to form the number of components corresponding to the number of characteristics having the number of changes is identified. The number of attributes for the number of composite materials for the number of characteristics having the number of changes based on the desired level of performance is changed.”
Testimony to (C) is Patent WO2015060960 (Filed on 18 Sep 2014): “Product Chemical Profile System”. The abstract describes as system that is able to pull together and query all levels of information about a product down to the chemistry level: “A computer-implemented system and method for obtaining product related information obtained from a plurality of different sources that is transformed into processed product data with a plurality of levels. Callouts and contexts are identified and a product-to-chemical continuum is generated by creating callout-context pathway segments between the plurality of levels of the processed product data based on the callouts and contexts identified and a transformed query request is generated used to traverse the product-to-chemical continuum through the callout-context pathway segments that span the plurality of levels. The product information that matches the set of context search parameters is extracted from the product-to-chemical continuum. The callout context pathway segments reduce processing resources and time needed to obtain the product information.”
These patents are a clear recognition of the relevance and importance materials modelling and a more integrated approach to engineering. The question remains however how to tear down barriers preventing its wider exploitation across the whole community. That is what the EMMC Roadmap, current and forthcoming Horizon2020 actions aim to address.
[i] A. Browning, “Utilization of Molecular Simulations in Aerospace Materials: Simulation of Thermoset Resin/Graphite Interactions,” Proceedings of AIChE Fall Annual Meeting, 2009.
[ii] Knox, C. K., Andzelm, J. W., Lenhart, J. L., Browning, A. R., & Christensen, S. (2010, December). High strain rate mechanical behavior of epoxy networks from molecular dynamics simulations. In Proc of 27th army science conf, Orlando, FL, GP-09.
Here is the executive summary of a new report on the economic impact of materials modelling, co-authored with Christa Court from MRIGlobal in the framework of the European Materials Modelling Council (EMMC) and the International Materials Modelling Board (IM2B). The full text as well as survey form is available here.
At the core of the report is an industry survey conducted during 2015 that provides corroboration for the indicators of research and development (R&D) process improvements found in earlier studies and new data relevant for quantitative economic analyses.
The survey is set in the context of an outline of metrics and methodologies that can be used to quantify the economic impacts of materials modelling from a variety of perspectives including R&D and industry stakeholders and society at large. At the micro-economic level, performance indicators include financial metrics such as net present value, return on investment (ROI), and internal rate of return. Where sufficient data are available, micro-economic analyses could be extended to a more in depth cost benefit analysis. Finally, macro-economic modelling methodologies can be used to model the wider impacts of the integration of materials modelling into the production function of various industries. Since materials modelling is a potentially disruptive technology, macro-economic impact assessment will likely require dynamic simulation models, which are scenario specific and necessitate someone with a high level of both problem domain knowledge and modelling domain knowledge.
Research impact is reviewed briefly based on bibliometrics, case studies, peer review, and economic analysis  using evidence gathered for a previous report  as well as the recent UK Research Excellence Framework , which includes 15 cases involving materials modelling.
The study also investigates how materials modelling impacts the industrial R&D process and outlines the value and potential of materials modelling for industrial research and innovation, competitiveness, and profitability using examples from materials industries based on recent Integrated Computational Materials Engineering studies and a Computer-Aided Drug Design study, which demonstrated the usefulness of defining a performance metrics for a modelling function in an industrial R&D organisation.
The survey analysis was based on information provided by 29 companies covering a wide range of sizes and industry sectors and an even distribution in terms of types and scales of modelling. The qualitative benefits identified in the responses were categorised into the following Key Performance Indicators: More efficient and targeted exploration; Deeper understanding; Broader exploration; R&D strategy development; Source of property data; Trouble shooting; Performance optimisation; Intellectual property protection; Value chain benefits; Improved communication and collaboration between R&D and production; Upscaling and market introduction as well as marketing benefits.
On a quantitative level about 80% of companies reported innovation accomplishment, 60% cost savings, 35% job creation, and 30% revenue increase due to materials modelling. A wide variety of project sizes are represented, with total materials modelling investment (covering staff, software and hardware) ranging from €45K to €4M (average €1M, median €½M). Staff was the largest cost factor: the ratio of staff costs to the median cost of software and hardware, respectively, is 100/20/6. Cost savings due to the materials modelling project ranged from €100K to €50M (average €12M, median €5M). The ROI, determined by the ratio of revenue generated and investment in modelling, ranged from 2 to 1000. Removing the largest and the smallest values yields an average ROI of 8. A trend for ROI to grow more than linearly with investment in modelling was found.