Raw feature counts from metagenomic sequencing are not directly comparable across samples due to variations in sequencing depth and initial DNA input. Absolutifier addresses this by converting raw counts into absolute abundances.
The tool works by calculating a sample-specific scaling factor based on the total DNA weight (concentration × volume) present before sequencing. This scaling factor is used to absolutify the raw counts, allowing for valid simulation of scale across samples.Absolutifier estimates uncertainty in these absolute abundances by adding a pseudocount to the observed counts and then drawing Monte Carlo samples from a Dirichlet distribution and provides a measure of error (95% confidence intervals for each feature).
Absolutifier works with any type of biological features: taxa, genes, functional categories, etc.
Clone or download the repository and install it using pip:
pip install .Optionally, use a virtual environment:
python3 -m venv venv
source venv/bin/activate
pip install .Run Absolutifier with:
absolutifier \
--counts data/PE_fastq/counts.csv \
--meta data/PE_fastq/metadata.csv \
--output output_pe_absolute.csv \
--volume 500 \
--fastq_folder data/PE_fastq \
--suffixes _R1 _R2 \
--extension .fastq| Flag | Description |
|---|---|
--counts |
CSV file with sample rows and feature columns (e.g., sample_id,feature_1,...) |
--meta |
CSV file with sample_id and DNA_conc columns |
--volume |
DNA volume (microL) before sequencing (fixed for all samples) |
--output |
Output CSV file with absolute abundances |
--fastq_folder |
(Required) Folder with FASTQ/FASTA files |
--extension |
(Optional) File extension (default: .fastq) |
--suffixes |
(Optional) List of suffixes for filtering file names (e.g., _R1 _R2) |
--singleton |
(Optional) Additional single FASTQ file(s) to include |
--error_bars |
(Optional) Flag to calculate 95% confidence intervals with Monte Carlo sampling |
--mc_samples |
(Optional) Number of Monte Carlo samples for error bars (default: 1000) |
--alpha |
(Optional) Dirichlet prior for the Bayesian error model (default: 0.5) |
--plot |
(Optional) Flag to generate visualization plots of the results |
--top_features |
(Optional) Number of top features to include in plots (default: 20) |
--plot_format |
(Optional) Image format for plots (png, pdf, svg; default: png) |
--figsize |
(Optional) Figure size for plots (width height; default: 12 8) |
counts.csv:
sample_id,feature_1,feature_2,feature_3
sample_1,0,5,2
sample_2,2,0,10
sample_3,100,4,0metadata.csv:
sample_id,DNA_conc
sample_1,22.5
sample_2,31.1
sample_3,29.3- Scaling factor: Calculates scaling factors by comparing DNA input to total sequenced base pairs.
- Robust Error Propagation: Computes 95% confidence intervals using a Bayesian model with Monte Carlo sampling.
- Comprehensive Outputs: Generates consolidated tables with counts, absolute abundances, confidence intervals, and scaling factors.
- Publication-Ready Plots: Creates high-quality bar plots, heatmaps, and confidence interval comparisons.
- Flexible Input: Supports paired-end, single-end, and mixed FASTQ/FASTA files.
Feel free to open issues or contribute to development!