Chemeia: A synergistic AI integrated architecture for augmenting high value dark-data.

跨科学研究和开发(R&D)环境的数据管理是确保内部实验和外部出版物的结果得到有效组织和整合以供进一步重复使用、重新解释和参考的关键过程。它支持与大数据分析相关的许多广泛的下游活动，例如药物发现信息学中的化合物选择或先进材料制造中的工艺优化。这些高价值的活动吸引了广泛的竞争性算法软件解决方案(例如，Reaxsys、Invenity Path Analysis、MaterialUniversal、EBI、IBM Watson)-所有这些解决方案都旨在最大化现有内容，并且往往依赖数据精选来产生它。当前的科学数据精选方法完全不够：高达85%的科学研究可能被浪费，90%的科学家认为存在可重复性危机(Munafo_Et Al_2017)-这是一个耗资数十亿美元的全球问题。这在很大程度上是因为这项任务的复杂性，需要专家手动评估文本和数据--扩展起来速度太慢，成本太高。现代人工智能(AI)方法仍然太不准确；它们也没有解决稀疏数据完成-这是集成实验数据时常见的问题。在这里，我们描述了Chemeia(‘Chem-ee-a’)，这是一个全新的通用数据管理解决方案，它结合了两种以最先进的人工智能为中心的技术，将静态的、预先构建的、实验数据库变成更完整、可靠和持续监控的资源，用于这些高价值的研发环境。BioRelate和IntelLegens是两家高科技公司，专门应用新的人工智能技术来解决为研发优化数据管理方面的众所周知的问题。BioRelate开发了Galaxy AI，这是一个统一了数据挖掘、自然语言处理和深度学习的创新平台，用于管理文本数据。它已被用于为专注于早期药物发现的大型制药和生物技术公司提供服务。IntelLegens开发了炼金术，它建立在现有的数字精选资源(稀疏数据)的基础上，通过使用人工智能来预测排名的未知数据点和不确定性。市场上没有其他解决方案可以复制AlChemite针对大型稀疏数字数据的低成本建模、预测、错误检测和参数优化。今年早些时候，两家公司都独立提高了客户数据库中内容的数量和质量，并意识到每种方法都受到其验证新结果和预测的能力的限制。将这两种技术结合在一起，使它们协同发挥作用，结合优势并解决验证问题，将产生一个能够产生更详细和准确数据的系统。“

"Data curation across scientific research and development (R&D) environments is a critical process for ensuring results from in-house experiments and external publications are effectively structured & integrated for further reuse, reinterpretation & reference. It supports a number of broad downstream activities associated with big data analytics, such as compound selection in drug discovery informatics or process optimisations in advanced materials manufacturing. These high value activities attract a broad range of competitive algorithmic software solutions (e.g. Reaxsys, Ingenuity Pathway Analysis, MaterialUniverse, EBI, IBM Watson) - all geared towards maximising existing content & are often reliant on data curation to produce it.Current methods of scientific data curation are wholly insufficient: up to 85% of scientific research is potentially wasted and 90% of scientists believe there is a reproducibility crisis (Munafo _et al_, 2017) - a global issue costing billions. This is largely down to the complexity of the task requiring experts to manually assess text & data - too slow and costly for scaling. Modern Artificial Intelligence (AI) approaches are still too inaccurate; nor do they address sparse-data completion - a common problem when integrating experimental data. Here we describe Chemeia ('Chem-ee-a'), an entirely new solution to general data curation that combines two state-of-the-art AI-centric technologies for turning static, pre built, experimental databases into more complete, reliable, and continuously monitored resources for use in these high-value R&D environments.Biorelate and Intellegens are two high-technology companies specialising in applying novel AI techniques to solve well known problems in optimising data curation for R&D. Biorelate have developed Galactic AI, an innovative platform unifying data mining, natural language processing & deep learning for curating textual data. It has been used to deliver for large pharmaceutical and biotech companies focusing on early stage drug discovery. Intellegens have developed Alchemite, which builds on existing numerical curated resources (sparse data) by using AI to predict ranked unknown data-points and uncertainties. No other solution on the market can replicate Alchemite's low cost modelling, prediction, error detection & parameter optimisation for big, sparse, numerical data.Earlier this year, both companies independently improved the volume and quality of content in a customers database and realised that each approach was limited by its ability to validate new results & predictions. Combining these two technologies so that they function synergistically, combining strengths & solving validation issues will result in a system that is capable of generating more detailed and accurate data."