NMF Clustering: Accessible NMF-based Clustering Utilizing GPU Acceleration.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Non-negative Matrix Factorization (NMF) is crucial for reducing high-dimensional gene expression data into interpretable metagenes.
• The computational intensity of NMF has limited its application on large datasets like single-cell RNA sequencing (scRNA-seq) data.

Purpose of the Study:

• To implement a computationally efficient NMF clustering algorithm suitable for large-scale gene expression datasets.
• To accelerate NMF analysis for RNA-seq and scRNA-seq data, making it more practical for researchers.

Main Methods:

• Developed a GPU-accelerated NMF clustering method utilizing CuPy and Message Passing Interface (MPI).
• Integrated the NMF clustering tool into the GenePattern gateway for accessible, web-based analysis.
• Enabled multi-step analysis pipelines on high-performance computing (HPC) clusters for reproducible research.

Main Results:

• Achieved a reduction in computation time by up to three orders of magnitude compared to traditional methods.
• Made NMF clustering analysis of large RNA-seq and scRNA-seq datasets computationally practical.
• Provided free public access to the NMF clustering tool via the GenePattern gateway.

Conclusions:

• The GPU-accelerated NMF clustering significantly enhances the feasibility of analyzing large omics datasets.
• The GenePattern integration democratizes access to advanced bioinformatics tools for non-programmers.
• This advancement supports reproducible in silico research in genomics and transcriptomics.