Computational genomics software engineer at UC Santa Cruz (labs of Benedict Paten and Karen Miga), working in the Human Pangenome Reference Consortium. I build fast, accurate software for genome assembly, pangenome graphs and long-read variant calling, in C++, Python and Mojo.
- pgphase: pangenome graph-aware variant calling and phasing for long reads. On HG002 PacBio HiFi: 55% fewer small-variant errors than DeepVariant, and 99.08% of reads phased correctly against 94.41% for HiPhase.
- Shasta: core developer of the de novo assembler for Oxford Nanopore reads behind NIH CARD's Napu pipeline.
- sporos: seeding library for read mappers and aligners (minimizers, syncmers, MEMs, MUMs, SMEMs). SMEMs 4.8× faster than minibwa; MEMs and MUMs 4.2–43× faster than MUMmer4, with identical output.
- plegma: fast partial-order multiple sequence alignment. In Shasta2's local assemblies it leaves 40% fewer consensus errors than abPOA, and in pgphase it makes 7× fewer false heterozygous calls.
- karan: exact pairwise sequence alignment for the CPU and the GPU, 1.3–7× faster than the next fastest exact aligner on real reads.
- dromos: reads vg pangenome graph alignments at disk speed.
- periodos: tandem repeat discovery. It scans human chr1 in about 18 s and, on HG002, recovers 99.9% of TRF's loci and 100% of GIAB's perfect short tandem repeats.