Biology-aware machine learning methods for characterizing microbiome genotype and phenotype

项目总结

PROJECT SUMMARY The Mirarab laboratory designs computational methods for answering biological and biomedical questions, fo- cusing on scalability and accuracy. These methods span several areas (e.g., microbiome profiling, multiple sequence alignment, and phylogenomics), and a common thread among them is evolutionary modeling. More recently, many of the developed methods are based on machine learning. The lab has developed scalable and accurate methods for reconstructing evolutionary histories (i.e., phylogenies) and using these histories in down- stream biomedical applications. Methods developed by this lab (e.g., ASTRAL, SEPP, DEPP) are at the fore- fronts of modern genome-wide phylogenetics. While the lab has previously focused more on inferring species histories, through an MIRA grant, it has shifted its focus to developing methods for microbiome analyses, which pose their a unique set of challenges. As part of the MIRA application, the Mirarab lab will focus on designing, testing, and applying improved methods for statistical analyses of microbiome data. These methods will target two questions. (i) Profiling: What organisms constitute a given sample? (ii) Association: How are samples different in their organismal composition, and how do these differences connect to measurable characteristics of their environment? While both questions have been subject to considerable research, many computational challenges remain, providing an opportunity for better methods to make a significant impact. Instead of focusing solely on new algorithms, the lab will also work on building better reference datasets and combining data from multiple sources. Thus, the project aims to harness the unprecedented computational power, large available datasets, and recent advances in machine learning to improve state-of-the-art dramatically. The project will not use off-the-shelf machine learning methods in a black-box fashion. Instead, it develops methods that incorporate biological knowledge (e.g., of the evolutionary relationships) into machine learning methods in a principled biologically-motivated fashion. Within the context of the MIRA award, this supplementary request is to purchase a computing server. The server will enable the lab to take advantage of the unprecedented level of genomic data available today to build machine learning methods that are trained on a much more representative set than existing methods. Thus, the extra computational power will not be just in the service of making analyses faster: it will enable using large datasets for training that could not be otherwise used.