SBIR Phase I: Semantic Link Association Prediction for Phenotypic Drug Discovery

这个小企业创新研究（SBIR）项目的更广泛的影响/商业潜力是开发一流的预测表型分析（PPP）软件工具，这将提高临床前药物发现过程的效率和有效性。最近对1998年至2008年间FDA批准的药物进行的一项研究表明，现在大多数一流药物都是从表型筛选而不是传统的基于靶标的筛选中获得的。然而，目前严重缺乏计算和数据工具，可以将大量传统的基于分子的数据与现在生成的同样大量的表型数据连接起来。PPP工具对这些复杂的、多方面的数据进行整合和解读，极大地提高了制药公司发现新药和有效药物的能力。每个新处方药批准的成本估计为25.6亿美元——将临床前药物发现过程缩短一周的经济影响估计将为制药业节省1.08亿美元的成本，创造巨大的金融机会。该工具旨在提高进入临床试验的药物的数量和质量，从而使人们能够获得价格更经济的药物。该SBIR一期项目建议开发一个概念验证PPP软件工具，该工具将各种公开可用的分子和表型数据源整合到图形用户界面中，允许发现新的作用机制，并从表型分析中识别目标。该项目的主要障碍将是这些高度异构的数据集的集成和基于证据的路径模式的识别。语义技术和领域专业知识将应用于该应用程序，以克服这些数据集成和预测挑战。实现原型PPP工具目标的计划包括：1)为表型数据源创建语义图，2)在表型数据中寻找循证路径模式，3)应用预测算法进行表型数据分析，以及4)开发用于评估和验证的图形用户界面。第一阶段的成功将产生一种可被制药公司用于评估和产品反馈的工具。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) project is the development of a first-in-class Predictive Phenotypic Profiler (PPP) software tool that will improve the efficiency and effectiveness of the pre-clinical drug discovery process. A recent study of drugs approved by the FDA between 1998 and 2008 shows that a majority of first-in-class drugs are now derived from phenotypic screens rather than traditional target-based screens. However, there is currently a severe lack of computational and data tools that can bridge the vast amounts of traditional molecular-based data with the equally vast amounts of phenotypic data now being generated. The PPP tool integrates and interprets this complex and multi-faceted data to greatly enhance the ability of pharmaceutical companies to find new and effective drugs. The estimated cost per new prescription drug approval is $2.56 billion - the economic impact of reducing the pre-clinical drug discovery process by just one week is estimated to result in a $108 million cost savings for the pharmaceutical industry, creating a large financial opportunity. This tool aims to enhance the number and quality of drugs that enter clinical trials, resulting in more economically priced medicines available to the population.This SBIR Phase I project proposes to develop a proof-of-concept PPP software tool that brings together a variety of publicly available molecular and phenotypic data sources into a graphical user interface, allowing for the discovery of novel mechanisms of action, and the identification of target(s) from phenotypic assays. The major hurdles of this project will be the integration of these highly heterogeneous datasets and the identification of evidence based path patterns. Semantic technologies and domain expertise will be applied to this application to surmount these data integration and prediction challenges. The plan to reach the goal of a prototype PPP tool includes: 1) Creating a semantic graph for phenotypic data sources, 2) finding evidence-based path patterns in phenotypic data, 3) applying predictive algorithms for phenotypic data analysis, and 4) developing a graphical user interface for evaluation and verification. Phase I success will result in a tool that can be used by pharmaceutical companies for evaluation and product feedback.