Sergey Koren, Ph.D.

Dr. Koren researches algorithms for the efficient analysis of large-scale genomic datasets with a focus on genome assembly. 

Genome assembly is the process of reconstructing a complete sequence from relatively short-range data generated by sequencing instruments, like a giant jigsaw puzzle with billions of pieces. It is sometimes considered routine due to the wide availability of short-read sequencing. However, no complex genome is truly complete. The challenge are repetitive regions within genomes which cannot be properly localized. Another limitation of current genome assembly software is that it represents every genome as one sequence. This is despite the fact that sequenced genomes are usually diploid, meaning they contain two copies of each chromosome, one from each parent. Trying to represent both copies as a single sequence results in a mosaic, as if you superimposed two images on top of each other, and leads to both data loss and errors in analysis. To mitigate these issues, the first step of an assembly project is often a laborious and year-long effort of inbreeding, to make the two copies as similar as possible. This is error-prone and often infeasible, due to long generation times (e.g. cattle and other agricultural species). 

Dr. Koren pioneered the use of newer noisy long-read data (such as generated by Pacific Biosciences (PacBio) and Oxford Nanopore (ONT) instruments) for high-quality assembly. This work has revolutionized bacterial genome assembly, giving complete and accurate sequences, something which previously required years of manual effort and has helped optimize algorithms at all steps of the assembly process. To assemble diploid genomes, Dr. Koren led the development of an approach which uses parental genomic information to identify sequences belonging to either the maternally or paternally inherited genome in the child to generate two complete sequences for a single individual. This approach is the current state of the art for accurately reconstructing a complete diploid genome. Most recently, Dr. Koren supervised a project to update the popular Canu assembler to optimally utilize newly available highly accurate (over 99% accurate) long reads. His research showed that, through several algorithms, data accuracy can be increased even further, essentially to perfection. These breakthroughs allowed the study of previously invisible human genomic regions, corrected errors in the current reference, and led to the first truly complete human genomic sequence. Dr. Koren’s work continues to build on this success so that complete and accurate genomes become routine.