CRISPR guide capture extraction: 10x GEM-X public lane

• Category: Public data

• Assay type: crispr

• Evidence level: Validation milestone

• Dataset: 10x GEM-X A375 CRISPR Guide Capture R2 sample

• Condition shown: Single-guide fixed-window guide-capture extraction with exact-slice baseline agreement.

• Primary report: docs/benchmarks/perturb_seq/README.md

What This Shows

• DotMatch can extract and count a fixed-window guide-capture signal against one observed guide target.

• The paired diagnostic fixture still checks side-level ambiguity handling.

What It Does Not Show

• This is not useful multi-guide assignment evidence.

• This is not guide-per-cell calling.

• Expression processing and perturbation-effect analysis remain outside DotMatch.

Comparator And Validation

Comparator semantics: The checked public lane uses a 10x CRISPR Guide Capture R2 subsample, one observed guide target, an observed fixed guide window, and a transparent exact-slice hash baseline.

Validation: The public gate requires synthetic pair-count diagnostics, public DotMatch k=0 and k=1 rows, zero validation mismatches, exact agreement with the k=0 baseline, and k=1 assignment count no lower than k=0.

Report Examples

• Perturb-seq/guide-capture benchmark report (markdown)

Raw Artifacts

• benchmarks/raw/perturb_seq.csv

• examples/perturb_seq/data/metadata.json

Reproduce Or Check

• make bench-perturb-seq-public

• make perturb-seq-public-gate