Machine learning and biology

点击翻译按钮获取中文摘要

A key challenge in biomedical research is integrating the massive amount of available omics data to better understand gene function. We developed GeneMANIA, a software tool to integrate multi-omic data and use machine learning to predict gene function based on similarity to genes of known function. A web server (http://genemania.org) enables users to make gene function predictions using 600 million gene-gene functional interactions from diverse omics data sets. GeneMANIA serves 10,000 users per month. We now propose to significantly expand the GeneMANIA algorithm and software to improve its utility for contemporary biomedical research.******Our main activity will be increasing GeneMANIA accuracy and scalability with novel machine learning methods. To enable GeneMANIA to scale to support the large number of functional interaction networks, we will update its algorithm using a novel combination of state-of-the-art machine learning techniques: deep learning and gene embedding. We will evaluate the scalability and accuracy of published and novel methods to identify the best one to improve the GeneMANIA framework.******Dissemination: We will leverage our large user base to facilitate dissemination, via collaboration, websites, publications, social media, conference presentations and training via the popular Canadian Bioinformatics Workshop training program.******Benefits: GeneMANIA benefits tens of thousands of biological researchers by helping them sift through large amounts of omics data to learn more about gene function. Our proposed work will vastly increase the value of the GeneMANIA technology for biomedical research. The ease of use of the GeneMANIA technology alleviates the need for specialized bioinformatics support for biologists and facilitates the use of very large genomics data already collected using massive government investment. This will lead to better access to analysis services, more efficient use of research funding, new scientific discoveries and new medical treatments.**