Welcome to DEBIAS-M! DEBIAS-M is a computational methods designed to identify and correct for processing bias and batch effects in microbiome data.
Every folder within this repository contains analyses to produce a different component within the results or plots directories. The different analyses contained within each folder are as follows:
| File/Folder | Description |
|---|---|
| CRC-simulations | Code to run and plot the simulutations synthetically varying different underlying parameters in the colorectal cancer prediction benchmark (Fig S5) |
| pseudocount-evaluation | Code evaluating the impact of pseudocounts and flooring as pre- and post-processing steps for DEBIAS-M (Fig S10a-c) |
| HIV-using-age | Code evaluating the use of different covariates during batch-correction on the HIV benchmark (Figs 2a, S3c) |
| regression | The regression benchmark for metabolite prediction (Figs 5c, S8) |
| regression-all-methods | The regression benchmark for the plot incorporating a wider array of batch-correction methods (Fig 5b) |
| Simulations | Implementing the simulations using synthetically generated data (Figs 3, S4, S6) |
| runtime-benchmark | Evaluates DEBIAS-M's runtime under various scenarios (Fig S10h,i) |
| convex-assumption | Generates visualizations describing the optimization space of DEBIAS-M, illustrating non-convex spaces |
| single-batch-cv | Testing DEBIAS-M as a single-study processor correcting for biases while only considering a single batch (Fig S9b) |
| study-weighting | Evaluates the impact of weighting the influence of each study's impact on the cross-batch loss based on each study's size (Fig R3-4) |
| loss-functions | Evaluates DEBIAS-M on some of the main benchmarks using a wide variety of loss functions (Fig S10f,g) |
| taxonomy-aggregation | Evaluates the impact of aggregating microbial features to different levels of taxonomy (Fig S10d,e) |
| new-cervix-carcinoma | Implementation of DEBIAS-M cervical carcinoma analyses, using a newly created combination of cervical datasets (Figs 2d, 6a,b) |
| within-control-comparison | Evaluates a version of Online DEBIAS-M in which similarity is only enforced within the controls of the training samples (Fig R2-2) |
| v1-DEBIAS-M-Analysis | Includes code for all remaining analyses (see nested README within this folder) |
| make-PR-csv.ipynb | Saves the table describing all auPRs in the main benchmarks (Supplementary Table 2) |