tangermeme's extract_loci now 10x faster: ATAC-seq locus loading drops from 3.64s to 0.35s
jmschreiber91 · x · 2026-10-11
jmschrei merged a PR speeding up extractloci in tangermeme, aiming to make it one of the fastest ways to load genomics data from disk onto a GPU for sequence-to-function modeling. FASTA windows are now read via memory-mapped .fai index and one-hot encoded with numba kernels across njobs threads, while bigWig reading is replaced by the new figwig project. On 167,750 ATAC-seq peaks and negatives with one bigWig at 8 threads, a call takes 0.35s versus 3.64s on main (0.09s vs 3.25s without signals), with outputs, errors and warnings verified identical across 11,039 calls.
More from Research
- DeepMind weather model nailed Hurricane Isaias track 11 days out — joao_gante · 2026-10-12
- Models trained on their own output lose rare knowledge first, drift into nonsense — TejasKumar_ · 2026-10-12
- Stanford Paper Uses CNN on Raw Prices to Predict S&P 500, Challenging Factor Tradition — mdancho84 · 2026-10-12
- Paradromics BCI lets clinical participant communicate via imagined speech — PeterDiamandis · 2026-10-12
- Did OpenAI limit solutions to protect academia? Debating decentralized, GitHub-speed research — IgorCarron · 2026-10-12
- CSCW2026 Keynote to Propose Expanding Research Agenda to Human-Agent and Agent-Agent Collaboration — merrierm · 2026-10-12