Translating data into knowledge: From biomedical sequences and text to disease prediction and prevention

自人类基因组测序以来产生的大量数据为理解常见复杂疾病背后的生物学过程和治愈这些疾病带来了希望。为了有效地利用数据并实现这些希望，必须开发出为特定生物医学需求精心定制的计算分析工具。总而言之，拟议研究的目标是开发生物医学数据的计算、智能和统计数据分析工具，并以支持对疾病的解释和预测的方式对复杂的生物系统进行建模。 在过去的几年里，我和我的团队与世界各地的知名研究人员合作，开发了几种高效的工具来解决当前生物医学中的突出问题。例如，我们开发了SHERLOC，这是迄今为止报道的最全面和最准确的系统，通过利用文本和序列数据来预测蛋白质在细胞内的位置。我们介绍了F-SNP--一种用于评估和预测小突变对基因组的功能影响的集成系统，称为单核苷酸多态(SNPs)。我们还提出了一种在已发表的文档中使用图像数据的新方法，作为通向目标明确的生物医学文本挖掘的一步。 在我们在有效利用和整合文本和序列数据方面取得的成功的基础上，我建议开发方法来识别科学出版物中文本和图像中的生物信息区，以及基因组上高度信息量的功能突变。然后，我计划将产生的工具专业化，并将其应用于大量与疾病相关的数据(特别是自闭症和癌症)，支持早期预测、干预和有效的药物设计。

The tremendous amount of data produced since the sequencing of the human genome raise hopes for understanding biological processes underlying common complex diseases and for curing them. To effectively use the data and realize these hopes, computational analysis tools - carefully tailored for specific biomedical needs - must be developed. In broad terms, the goals of the proposed research are to develop computational, intelligent, statistical data analysis tools for biomedical data, and to model complex biological systems in a way that supports explanation and prediction of disease. In the past few years, I and my group, in collaboration with prominent researchers throughout the world, have developed several highly effective tools that address eminent current problems in biomedicine. For instance, we developed SherLoc, the most comprehensive and accurate system reported to-date for predicting proteins' location within the cell, by utilizing text and sequence data. We introduced F-SNP - an integrated system for assessing and predicting the functional impact of small mutations on the genome, known as Single Nucleotide Polymorphisms (SNPs). We also proposed a new way to use image data within published documents as a step toward well-targeted biomedical text mining. Building on our proven success in effectively using and integrating text and sequence data, I propose to develop methods for identifying biologically informative areas in text and in images within scientific publications, as well as highly informative functional mutations on the genome. I then plan to specialize and apply the resulting tools to a wealth of disease-related data (specifically Autism and Cancer), supporting early prediction, intervention and effective drug-design.