BARCS Studio
Runs on this device Documentation

Beta-binomial analysis for modern CRISPR screens

Keep the full design.
Test the coefficient.

Analyze time, dose, ordered phenotype, treatment, and covariate-adjusted pooled screens without reducing the experiment to a high–low comparison.

logit(μgi) = xiTβg

Counts remain tied to full-library depth. Independent libraries—not reads—set the residual degrees of freedom.

CSV / TSV local worker downloadable results
β

Your coefficient results will appear here.

Count FASTQ guide k-mers or load a count matrix, add sample metadata, and BARCS will fit each guide locally.

  • CB2-derived rolling k-mer counting with reverse-orientation fallback
  • Mapping, representation, and count-distribution QC
  • Guide effects with Student-t inference
  • Dispersion and convergence diagnostics

What this web version preserves

01

Full-library denominators

Totals are computed before filtering, so subsetting guides cannot silently change the likelihood.

02

Sample-level inference

Residual degrees of freedom come from independent libraries and the chosen design matrix.

03

Visible boundaries

The app rejects rank-deficient designs and reports convergence, dispersion boundaries, and calibration scale.

04

R-reference parity

The same equations and decisions are tested against R. Small floating-point differences are expected; results are not promised to be bit-for-bit identical.

Input format

Column names may be comma- or tab-separated. Count sample columns must match metadata sample values exactly.

counts.csv

guide,gene,control,D0_A,D0_B,D7_A,D7_B
g1,GENE1,false,120,98,54,47
nt1,NTC,true,85,91,87,90

metadata.csv

sample,time,batch,total
D0_A,0,A,1824000
D0_B,0,B,1912500
D7_A,7,A,1768300
D7_B,7,B,1897100

library.csv

guide,gene,control,sequence
g1,GENE1,false,ACGTACGTACGTACGTACGT
nt1,NTC,true,TGCATGCATGCATGCATGCA

library.fasta

>g1 gene=GENE1
ACGTACGTACGTACGTACGT
>nt1 gene=NTC control=true
TGCATGCATGCATGCATGCA

FASTQ sample names are filenames without .fastq, .fq, or .gz, and must match metadata. The browser uses CB2-derived exact rolling k-mers: the first forward guide hit is counted, and the reverse complement is searched only when no forward hit exists. This completes CB2's stored-but-unreturned reverse-count path. Every library must use one guide length. Repeated guide sequences are excluded. For valid Student-t inference, independent libraries must outnumber fitted design columns.