RAFTS³G: an efficient and versatile clustering software to analyses in large protein datasets

Area of Science:

• Bioinformatics
• Computational Biology
• Data Mining

Background:

• Clustering methods are crucial for partitioning biological samples and reducing complexity in large datasets.
• Existing data mining tools often use greedy algorithms that can degrade biologically relevant information.
• Lack of standardized metrics and benchmarks questions the efficiency and reliability of current clustering methods.

Purpose of the Study:

• To introduce a novel approach for data mining in large protein sequence datasets.
• To develop a clustering method that minimizes the loss of biological information during group generation.
• To enhance accuracy and sensitivity in clustering across a wide range of sequence similarities.

Main Methods:

• Development of the Rapid Alignment Free Tool for Sequences Similarity Search to Groups (RAFTS³G).
• Implementation of an optimized, stringent algorithm for increased accuracy and sensitivity.
• Application of a binary search concept to grouped sequences for efficiency.

Main Results:

• RAFTS³G demonstrates superior performance compared to three main clustering methods.
• The tool achieves higher accuracy and sensitivity, even without an ideal clustering threshold.
• RAFTS³G effectively clusters millions of biological sequences, showcasing remarkable efficiency.

Conclusions:

• RAFTS³G offers an efficient clustering solution for large biological datasets, handling millions of sequences.
• The method balances the reduction of biological information redundancy with the creation of consistent groups.
• RAFTS³G minimizes processing time while maintaining a strong sensitivity/accuracy relationship.