Ompute

Ompute is a Python package for omics data missing value imputation, providing a unified interface to 22 individual and 10 Tribrid imputation models. Each model implements a reproducible classical, machine-learning, bayesian, or hybrid strategy to handle MCAR, MAR, and MNAR missing values in omics datasets.

Imputation Model Details

Individual Imputation Models

Tribrid (Hybrid) Models

Input and Experimental Design Structure

1. Input File (input.csv)

The package accepts only CSV files as input.
The input.csv file can contain both metadata columns (e.g., Gene, Protein, Protein Description) and data columns (quantitative features such as intensities or expression values) to be considered for imputation.
Only the data columns specified in the experimental design file (expdes.csv) will be subjected to imputation.

2. Experimental Design File (expdes.csv)

The expdes.csv file defines the structure and mapping between metadata and data columns for imputation.
It guides the imputation pipeline by indicating which columns in the input file correspond to biological metadata and which correspond to numeric data values.

Example:

3. Notes

• Both input.csv and expdes.csv must have matching column names for imputation and are case-sensitive.
• Missing values (NaN / 0 / blank) in data columns are imputed by the selected model.
• Other metadata columns remain same and reflected as it is conserving the original file structure.
• The package expects all input files in comma-separated CSV format with headers.

Usage

Each model can be executed directly as a function once the package is installed.

import ompute

# Example 1: Run an individual model
ompute.knn(i="data/input.csv", d="data/expdes.csv", o="/results")

# Example 2: Run a Tribrid model
ompute.trb1(i="data/input.csv", d="data/expdes.csv", o="/results")

# Example 3: Get imputed data as a pandas DataFrame
df = ompute.mean(i="data/input.csv", d="data/expdes.csv", o="/results", return_df=True)