Jeremy Buhler

Research

Advances in high-throughput DNA sequencing have led to a proliferation of sequence-based methods for genomics, metagenomics, transcriptomics and other large-scale molecular surveys at levels from individual cells, to organisms, all the way up to entire ecosystems. The challenge now is to efficiently aggregate, integrate, and search these data sets to make new discoveries and connections.

Jeremy Buhler's research focuses on developing algorithms and tools for large-scale computational analyses of biological sequences. He works to improve and generalize the heuristics at the heart of highly scalable computational tools for genomic, metagenomic, transcriptomic, and proteomic analysis, using ideas from probabilistic and randomized algorithms. These ideas have produced principled strategies to accelerate search in databases of sequences and probabilistic sequence models, as well as pattern discovery challenges such as motif finding.

Building computational infrastructure for next-generation sequence analyses also presents challenges of hardware and software design. By constructing accelerators for these analyses on parallel computing devices such as field programmable gate arrays (FPGAs) and graphics processors (GPUs), Professor Buhler exposes and addresses challenges to using these devices effectively. Recent work includes throughput-optimized mappings of dynamic programming recurrences to systolic arrays, design of deadlock-free dataflow applications in the presence of data filtering, and efficient mapping of work ensembles to threads on GPUs for irregular applications such as short-read aligners.