A Machine Learning Approach to Identify C Type Lectin Domain (CTLD) Containing Proteins.

Area of Science:

• Biochemistry
• Bioinformatics
• Genomics

Background:

• Lectins are proteins that bind carbohydrates and play diverse biological roles.
• C-type lectins (CTL) are involved in immunity and cell signaling, possessing a carbohydrate recognition domain (CRD).
• Identifying C-type lectin domain containing proteins (CTLD) is challenging due to low sequence homology.

Purpose of the Study:

• To develop a machine learning tool for accurate identification of CTLD.
• To overcome limitations of homology-based methods for CTLD detection.
• To discover novel CTLD within sequenced genomes.

Main Methods:

• Utilized machine learning, specifically Linear SVC classifier from Python's sci-kit library.
• Trained the model using dipeptide and tripeptide compositions of known CTLD and non-CTLD sequences.
• Optimized model parameters and dataset composition for improved performance.

Main Results:

• Initial model achieved precision of 0.92, recall of 0.91, and MCC of 0.82 on an external test set.
• Fine-tuning parameters and dataset improved performance to precision of 0.99, recall of 0.99, and MCC of 0.96.
• Successfully identified new CTLD in hypothetical regions of the human genome.

Conclusions:

• The developed machine learning tool accurately identifies CTLD.
• This tool enhances the discovery of C-type lectin domain containing proteins.
• The tool is accessible on a local server for research use.