ompute 1.0.0

Python package for Biological Omics Data Missing Value Imputation

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

Category	Model Name
MCAR	Mean Imputation
MCAR	Median Imputation
MCAR	Mode Imputation
MCAR	Hot-Deck Imputation
MCAR	KNN Imputation
MCAR	MissForest Imputation
MCAR	PPCA / BPCA
MCAR	SVD / SoftImpute
MCAR	Expectation-Maximization (EM)
MAR	MICE / Iterative Imputer
MAR	Linear / ElasticNet Regression Imputation
MAR	Random Forest Regression Imputation
MAR	Local Least Squares (LLS / LSA)
MAR	SoftImpute (Matrix Completion)
MAR	MissMDA (PCA Iterative)
MAR	Bayesian Regression (DPR)
MAR	Autoencoder Imputation
MNAR	MinDet (Minimum Detection Limit)
MNAR	MinProb (Minimum Probability)
MNAR	Perseus Downshifted Normal
MNAR	Truncated Normal Imputation
MNAR	QRILC (Quantile Regression Imputation of Left-Censored)

---

Tribrid (Hybrid) Models

Model	MCAR Component	MAR Component	MNAR Component
trb1	KNN Imputation	MissForest Imputation	QRILC
trb2	KNN Imputation	Random Forest Regression	QRILC
trb3	Mean Imputation	MICE / Iterative Imputer	MinProb
trb4	Median Imputation	Linear / ElasticNet Regression	Perseus Downshifted Normal
trb5	Mode Imputation	MissForest Imputation	MinDet
trb6	Hot-Deck Imputation	Local Least Squares (LLS / LSA)	Truncated Normal Imputation
trb7	KNN Imputation	MissForest Imputation	Perseus Downshifted Normal
trb8	SVD / SoftImpute	MissMDA (PCA Iterative)	QRILC
trb9	PPCA / BPCA	MICE / Iterative Imputer	MinProb
trb10	KNN Imputation	Random Forest Regression	MinDet

---

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:

Column Name	Group	Replicate
Control	Control	1
Control	Control	2
Control	Control	3
TreatmentA1	Test1	1
TreatmentA2	Test1	2
TreatmentA3	Test1	3
TreatmentB1	Test2	1
TreatmentB2	Test2	2
TreatmentB3	Test2	3

---

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)