🍹 bartab: Fitness from pooled competition assays

Using bartab v0.0.10.post1 | Documentation | Tutorial on analysis principles

Infer the competitive fitness of barcoded strains from next-generation sequencing of pooled growth experiments.

Upload your count table, sample sheet, and barcode sheet, then click Calculate fitness.

1️⃣ Input tables

Three tables are required. You can upload CSV, TSV, or XLSX files, or try one of the example datasets below.
Examples with synthetic data

🧮 Count table

One row per barcode per sample. Must contain:

  • a column of barcode/strain identifiers (matching your barcode sheet)
  • a column of sample identifiers (matching your sample sheet)
  • a column of read or UMI counts
Counts column

📶 Barcode information

One row per unique barcode. Must contain:

  • a column of barcode identifiers

Optionally: any metadata about strains (gene targets, constructs, etc.). These will be carried through to the output.
Barcode identifier column

🧪 Sample sheet

One row per sample (sequencing library). Must contain:

  • Sample ID: unique identifier matching the count table
  • Culture ID: biological replicate identifier. Samples from the same culture share this label
  • Timepoint: numeric timepoint values. The earliest timepoint is treated as $t_0$.

For spike-in normalisation: no extra columns needed. Just include your spike-in barcode in the count table and barcode sheet.

For growth-based normalisation: add a column of OD600, CFU/mL, or generation counts measured at each sample.

For dose-response analysis: add a column of inducer/drug concentrations. Samples with concentration = 0 are treated as uninduced controls.

For adaptive-volume sampling (if you took different volumes from each sample): add a column of sampled volumes.
Timepoint column
Concentration column
Individual sample ID column
Culture / biological replicate column
Volume column (if using)
Growth column (if using)

2️⃣ Control strains

  • Reference (WT): the strain relative to which all fitness values are calculated. Fitness = 1 by definition.
  • Spike-in: a non-growing strain (e.g. heat-killed or plasmid-only) added at a fixed concentration before library preparation. Used to infer how much the reference strain has expanded between timepoints, removing the need for growth measurements. Leave blank if using growth measurements instead.
Reference (WT) barcode name
Spike-in barcode name (if using)

3️⃣ Analysis options

  • Culture expansion: choose Spike if you have a non-growing spike-in control, or Growth if you have OD600/CFU measurements.
  • Pseudocount: added to all counts before log transformation to avoid log(0).
  • Analysis mode: choose Single concentration for standard fitness screens, or Dose response if your sample sheet contains a concentration column. Dose response fitting uses a 2-parameter Hill model to estimate the IC₅₀ and maximum effectfor each barcode.
Culture expansion uses:

Plotting options
Analysis mode

4️⃣ Results

Fitness values are estimated by weighted least squares regression of the log-ratio of each barcode against the reference strain, using the spike-in or growth measurements as the x-axis.

Key output columns:

  • For single concentration:
    • fitness: relative fitness ($w_i / w_{wt}$). Values < 1 indicate growth disadvantage; > 1 indicates advantage.
    • fitness_low / fitness_high: 95% confidence interval bounds.
    • slope_p: p-value for the slope being different from 0 (i.e. fitness ≠ 1).
  • For dose-response:
    • log_ic50 (log₁₀ concentration at 50% inhibition) and log_ic50_p.

Results and the full annotated dataset (.h5ad) can be downloaded below.