Bio-renewable Formulation Information and Knowledge Management System

该项目将建立一个示范信息和知识管理系统（IKMS），以促进在配方产品中使用可再生生物质的新化学材料和替代化学材料的创新。IKMS将使原料的简单转化中的功能成分能够更快地识别，并为特定功能推荐最佳原料。如果成功，它将修复供应链中生物基可再生材料作为配方产品功能成分的断裂，为商业合作伙伴创造巨大的商业利益，并在传播和进一步发展之后，为英国生物基材料部门和配方产品业务整体创造巨大的商业利益。该演示器将专注于寻找生物表面活性剂的创新，并将通过首次以激进的方法集成多个IT工具来制定产品设计，并成为首个在化学使用行业供应链中应用的此类工具，从而实现自身的创新。该系统的目标是将现有数据与从实验测量中恢复的新数据进行整理和管理，并使用这些数据来更新搜索工具应用的模型。为此目的，将开发一种自动数据驱动的建模工具，并将其集成到系统中。随着数据的增加和模型的改进，选择算法的性能将随着所选成分和候选配方满足下游商业化标准的机会而提高。值得注意的是，这里使用的建模方法完全不同，但与TSB资助的ICT项目101508下开发的建模方法是互补的，这些建模方法是基于物理而不是数据驱动的，并且将为针对子集过滤标准快速选择新化学物质提供强大的能力，并提供机制见解来锐化这些过滤器，以获得更高的精度和更好的实验分析设计。为了实现其目标，该项目将扩展101508信息模型，并添加一个存储库来存储配方信息（组合和组装）和属性数据（实验和计算），以补充原料和转换存储库。信息模型和存储库需要具有化学智能，使用易于扩展的RDF和三重存储技术，并结合语义搜索功能以促进集成。建模工具将使用现代机器学习方法进行调整和实现，以发现成分结构和属性之间的数学关系，以及配方组成和装配与应用性能之间的数学关系。模型将建立在项目期间创建的数据之上，并添加到101508模型存储库中。用于列举成分选项（来自原料和化学转化过程）的101508工具将扩展到列举配方（来自成分和组装过程）。枚举工具将与使用多样性或化学结构/配方组成/装配属性模型的全球多目标搜索工具耦合，以有效地探索组合成分/配方空间。我们还将开发工具，以最小的开销来帮助维护和发展IKMS。这些包括语义搜索和从文献和其他可用资源中提取适当数据的半自动提取，以及本体集成和不存在这些的本体的半自动构建。为了演示该系统如何在实践中工作，将制作101508年鉴定的新型生物表面活性剂并测量其性能，选定的子集制定和评估，数据和衍生模型用于驱动生物表面活性剂选择和配方优化的另一个周期。

The project will build a demonstration information and knowledge management system (IKMS) to facilitate innovation withnew and replacement chemical materials from renewable biomass in formulated products. The IKMS will enable functionalingredients from simple transformations of feedstocks to be identified more quickly and recommend the best feedstocks fora particular function. If successful, it will repair a disconnection in the supply chain for exploitation of bio-based andrenewable materials as functional ingredients in formulated products, creating significant business benefit to the commercial partners and, following dissemination and further development, to the UK bio-based materials sector andformulated products businesses as a whole. The demonstrator will focus on a search for bio-surfactant innovations, and willbe innovative in itself by both integrating several IT tools for the first time in a radical approach to formulated product designand by being the first of its kind to be applied across a chemical using industry supply chain.The ambition of the system is that it will collate and manage existing data with new data recovered from the experimentalmeasurements and use this to update the models applied by the search tools. An automated data-driven modelling tool willbe developed and integrated into the system for this purpose. As data is added and as models are improved, theperformance of the selection algorithms will improve along with the chances that the selected ingredient and formulationcandidates will meet downstream commercialisation criteria. It is important to note that modelling methods used here arequite different but complementary to those to be developed under the TSB funded ICT project 101508, which are physicsbasedrather than data-driven, and will provide powerful capability for fast selection of novel chemistries against a subset offilter criteria and provide mechanistic insights to sharpen these filters for better precision and better experimental assaydesign.To achieve its objectives, the project will extend the 101508 information model and add a repository to store formulationinformation (composition and assembly) and property data (experimental and computed) to complement the feedstock andtransformation repositories. The information model and repository will need to be chemically intelligent, use readilyextensible RDF and triple store technologies, and incorporate semantic search capabilities to facilitate integration.Modelling tools will be adapted and implemented using modern machine learning methods to find the mathematicalrelationships between ingredient structure and properties, and between formulation composition and assembly withapplication performance. The models will be built on data created during the project and added to the 101508 modelrepository. The 101508 tools for enumerating ingredient options (from feedstocks and chemical transformation processes)will be extended to enumerating formulations (from ingredients and assembly processes). The enumeration tools will becoupled to a global many-objective search tool using diversity or chemical structure/formulation composition/assembly -property models for efficient exploration of the combinatorial ingredient/formulation space.We will also develop tools to help maintain and grow the IKMS with minimal overhead to future projects. These includesemantic search and semi-automated extraction of appropriate data from literature and other available resources, and forontological integration and semi-autonomous building of ontologies where these do not exist.In order to demonstrate how this system will work in practice, novel bio-surfactants identified in 101508 will be made andtheir properties measured, a selected sub-set formulated and evaluated and the data and derived models used to driveanother cycle of bio-surfactant selection and formulation optimisation.