Next generation Text Mining in Drug Discovery

从文本文档中提取有趣的和重要的模式是生物化学科学知识发现的下一代浪潮。生物医学文献中的自由文本包含大量关于小分子及其靶标的信息，这些信息目前尚未存储在生化知识库中。这些信息可用于识别和建立药物-基因关联、化学和生物毒性甚至药物不良反应的特定特征。嵌入方法的最新进展在若干生物医学和临床任务中显示出有希望的结果。在生物医学记录上进行的文本分类提出了具体的挑战，包括数据集不平衡、拼写错误、缩写或语义模糊。目前最先进的方法将深度学习应用于任务，主要是卷积神经网络（CNN）、循环神经网络（RNN）、双向长短期记忆（Bi-LSTM）和BERT （Devlin等人，2019；Wolf等人，20）。该项目将通过优化命名实体识别（NER）程序和应用新颖的机器学习策略来生成您自己的语义词典，从而为Exscientia现有的文本挖掘平台做出贡献。它将在项目期间获得探索和人工智能技术团队的专业知识，以提供建议/支持。

Extracting interesting and non-trivial patterns from text documents is the next-generation wave of knowledge discovery in biochemical sciences. Free text resident in biomedical literature contains a wealth of information about small molecules and their targets that is not currently stored in biochemical knowledgebases. This information can be exploited to identify and build specific signatures for drug-gene associations, chemical and biological toxicity and even adverse drug effects.Recent advances in embedding methods have shown promising results for several biomedical and clinical tasks. Text classification performed on biomedical records poses specific challenges including dataset imbalance, miss-spellings, abbreviations or semantic ambiguity. Current state-of-the-art approaches apply deep learning to the task, mainly convolutional neural network (CNN), recurrent neural network (RNN), bi-directional long short term memory (Bi-LSTM), and BERT (Devlin et al.,2019; Wolf et al.,20).This project will contribute towards Exscientia' existing text mining platform by optimising named entity recognition (NER) procedures and applying novel machine learning strategies to generate your own semantic lexicon. It will have access to expertise across Discovery and AI technology teams to advise/support during the project.