Compression and k-Mer Based Approach for Anticancer Peptide Analysis.

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Anti-cancer peptide (ACP) sequence classification is vital for advancing cancer treatments.
• Current deep learning models for ACP classification demand extensive data and parameters.
• Existing compression methods may overlook crucial local sequence information.

Purpose of the Study:

• To develop a novel, efficient, and parameter-free method for ACP sequence classification.
• To preserve fine-grained amino acid context lost in traditional compression techniques.
• To provide a viable alternative to computationally intensive deep learning models.

Main Methods:

• Integration of Gzip compression with an incremental k-mer strategy.
• Compression of individual k-mers and sequential building of subsequence compressions.
• Utilizing Normalized Compression Distance (NCD) and kernel-based embeddings for classification.

Main Results:

• Achieved state-of-the-art performance on breast and lung cancer ACP datasets.
• Outperformed deep neural networks and large language models in classification accuracy.
• Demonstrated effectiveness without requiring custom features or pre-trained models.

Conclusions:

• The proposed Gzip-based incremental k-mer method offers a practical and efficient approach to ACP classification.
• This parameter-free technique excels in low-resource environments, overcoming limitations of current methods.
• The approach preserves essential amino acid-level context for improved classification accuracy in cancer research.