Overlap Desk - notice

The app's agent prompt is derived from the agent skill "gtars" (@k-dense-ai/gtars, skill version
1.3) in the repository k-dense-ai/scientific-agent-skills by K-Dense Inc.
  https://github.com/k-dense-ai/scientific-agent-skills (skills/gtars)
The skill's front matter declares the MIT licence. No text of the skill is redistributed verbatim;
the prompt was rewritten for this app.

The in-browser analysis (overlapkit.js) is an independent JavaScript implementation of the interval
semantics of gtars 0.9.2 (databio/gtars; PyPI gtars 0.9.2): BED read as 0-based half-open, strict
half-open overlap, RegionSet.reduce() merging overlapping and book-ended intervals, jaccard(),
coverage(), overlap_coefficient(), any_overlaps() and count_overlaps(). No gtars, bedtools or scipy
code is included.

It was checked against gtars 0.9.2 (Python), bedtools 2.26.0 (merge, intersect -u, closest -d) and
scipy 1.15.3 (fisher_exact) on 300 random pairs of region sets - one to four contigs including an
unplaced contig, contigs only one set uses, duplicated and book-ended intervals, universes - with
no disagreement in 38,679 comparisons: set sizes, merged base pairs, Jaccard, both coverages, the
overlap coefficient, intervals with an overlap, every per-interval overlap count, the reduced
intervals, the nearest-interval gap (bedtools 2.26.0 reports this gap plus one), the universe 2x2
table, the odds ratio and all three Fisher p-values. The seeded shuffle null was replicated exactly
by a separate Python implementation on 120 cases, and on 100 pairs of independent random sets on
hg38 the shuffle test called 5% enriched at p < 0.05.

The universe test follows the design of LOLA:
  Sheffield, N. C. & Bock, C. (2016). LOLA: enrichment analysis for genomic region sets and
  regulatory elements in R and Bioconductor. Bioinformatics 32(4), 587-589.
  https://doi.org/10.1093/bioinformatics/btv612

Genome presets are the primary chromosomes (chr1-22 or chr1-19, X, Y, M) of UCSC's hg38, hg19,
mm10 and mm39 chrom.sizes files (hgdownload.soe.ucsc.edu/goldenPath/<assembly>/bigZips/), fetched
2026-09-27.

If this skill contributed to a publication, K-Dense asks that it be cited:
  Kassis, T., Agarwal, V., He, Y., Patel, D., & Brueckner, A. M. (2026). Scientific Agent Skills:
  A Library of Procedural Knowledge for Research Agents. arXiv:2609.00065.
  https://doi.org/10.48550/arXiv.2609.00065

The example intervals in example.js are illustrative, generated with a fixed seed from stated
made-up models, not measured and not taken from any database: promoter windows of TSS +/- 1 kb
around random positions on hg38 chr1-8 (some genes given a nearby second TSS), accessible sites
placed at about two thirds of promoters and at random, and TF peaks drawn from accessible sites;
two replicate peak sets sharing 78% of a set of random sites with jittered edges, and a universe
of their merged union plus other random peaks; and hg19 enhancers, a few of them past the hg38
ends of chr1 and chr9, against Ensembl-named windows.