HCS-hierarchical algorithm for simulation of omics datasets.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Omics data analysis using machine learning (ML) is challenged by small sample sizes and high dimensionality.
• Existing omics data simulators primarily focus on generating high-fidelity molecular marker data.
• There is a need for generalized omics data simulation to test feature selection and ML algorithms.

Purpose of the Study:

• To develop a generalized omics data simulation approach.
• To create datasets that mimic real data structures for algorithm testing.
• To enable the generation of contrast variables with specific correlation structures.

Main Methods:

• Proposed an algorithm for omics dataset reconstruction.
• Utilized hierarchical clustering of variables and principal components of clusters.
• Preserved the correlation structure of original data with reduced parameters.

Main Results:

• The algorithm successfully reconstructs omics datasets with high fidelity.
• It preserves the correlation structure using a reduced number of parameters.
• The method reproduces topological descriptors of the correlation structure well.

Conclusions:

• The developed method provides a valuable tool for testing feature selection and ML algorithms.
• It allows for the simulation of omics data with controlled correlation structures.
• The approach aids in understanding and improving analytical methods for omics data.