Protein sequence classification with improved extreme learning machine algorithms.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Proteomics

Background:

• Accurate protein sequence classification is crucial for drug development.
• Conventional classification methods are computationally intensive and time-consuming.
• There is a need for efficient protein sequence classification systems.

Purpose of the Study:

• To evaluate the performance of single-hidden layer feedforward networks (SLFNs) for protein sequence classification.
• To develop and assess enhanced SLFNs using ensemble techniques for improved efficiency and accuracy.
• To compare proposed methods against existing approaches using benchmark datasets.

Main Methods:

• Utilized extreme learning machine (ELM) and its optimal pruned variant (OP-ELM) as training algorithms.
• Constructed ensemble-based SLFNs by combining multiple SLFNs trained with ELM or OP-ELM.
• Employed a majority voting method for final category determination in ensemble models.
• Validated performance using datasets from the Protein Information Resource center.

Main Results:

• The proposed ensemble SLFNs, particularly with OP-ELM, demonstrate superior performance in protein sequence classification.
• The optimized methods significantly reduce classification time compared to conventional approaches.
• Experimental results confirm the effectiveness and priority of the developed algorithms.

Conclusions:

• Ensemble-based SLFNs, leveraging ELM and OP-ELM, offer an efficient and accurate solution for protein sequence classification.
• The developed system addresses the limitations of traditional time-consuming methods.
• This approach holds promise for accelerating the development of pharmacological products through improved bioinformatics tools.