Evolving Transcriptomic Profiles From Single-Cell RNA-Seq Data Using Nature-Inspired Multiobjective Optimization.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Transcriptomic profiling is crucial for post-genomic analysis.
• Single-cell RNA-seq (scRNA-seq) offers high resolution for gene expression studies.
• Existing computational methods for scRNA-seq data face challenges like high dimensionality and premature convergence.

Purpose of the Study:

• To propose and formulate an evolutionary multiobjective blind compressed sensing (EMOBCS) framework.
• To address limitations in existing algorithms for transcriptomic profiling from scRNA-seq data.
• To enhance the accuracy and efficiency of gene expression analysis at the individual cell level.

Main Methods:

• Developed an EMOBCS framework utilizing an artificial bee colony algorithm.
• Formulated two objective functions: chi-squared kernel score and Euclidean distance for gene expression profiles.
• Introduced a rank probability model and novel search strategies within a cooperative convolution framework.

Main Results:

• The proposed EMOBCS algorithm demonstrated superior or comparable performance against 14 other algorithms across 10 scRNA-seq datasets.
• Extensive experiments validated the effectiveness of the new method in evolving transcriptomic profiles.
• Time complexity, convergence, and parameter analyses confirmed the algorithm's robust properties.

Conclusions:

• The EMOBCS framework provides an effective solution for transcriptomic profiling using scRNA-seq data.
• The proposed method overcomes key challenges associated with high dimensionality and convergence in computational transcriptomics.
• This advancement contributes to a more accurate understanding of gene expression at the single-cell level.