On the use of QDE-SVM for gene feature selection and cell type classification from scRNA-seq data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for cell type identification, but technical variations complicate analysis.
• Gene selection (feature selection) is vital for identifying informative genes in scRNA-seq data.
• Existing scRNA-seq gene selection primarily uses filter and embedded methods; wrapper methods are underutilized and often clustering-based.

Purpose of the Study:

• To explore a classification-based wrapper method for scRNA-seq gene selection.
• To introduce a quantum-inspired differential evolution (QDE) algorithm wrapped with machine learning classifiers for this task.

Main Methods:

• A quantum-inspired differential evolution (QDE) algorithm was developed and wrapped with various machine learning (ML) classifiers.
• The QDE-ML approach was tested on twelve scRNA-seq datasets for cell type identification.
• The linear Support Vector Machine (SVM) classifier wrapped with QDE (QDE-SVM) was selected based on experimental results.

Main Results:

• QDE-SVM demonstrated superior performance in cell type classification compared to other QDE-ML combinations.
• QDE-SVM outperformed existing wrapper methods (FSCAM, SSD-LAHC, MA-HS, BSF).
• QDE-SVM achieved a high average accuracy of 0.9559, significantly higher than the 0.8292-0.8872 range of other methods.

Conclusions:

• Classification-based wrapper methods, like QDE-SVM, are effective for scRNA-seq gene selection.
• The QDE-SVM approach offers a promising advancement for accurate cell type identification in scRNA-seq studies.
• This method addresses the need for robust gene selection in complex transcriptomic data analysis.