Genetic source completeness of HIV-1 circulating recombinant forms (CRFs) predicted by multi-label learning

Area of Science:

• Virology
• Computational Biology
• Machine Learning

Background:

• Human Immunodeficiency Virus type 1 (HIV-1) exhibits vast sequence diversity due to recombination between different subtypes.
• Inter-subtype genomic recombinants can evolve into circulating recombinant forms (CRFs) with significant public health implications.
• Accurately identifying the genetic origins of CRFs and detecting novel subtypes is a complex challenge.

Purpose of the Study:

• To develop and validate a multi-label learning algorithm for predicting the complete set of subtype sources for HIV-1 CRFs.
• To assess the algorithm's capability in identifying emerging subtypes and their genetic components.
• To predict the chronological order of CRF emergence.

Main Methods:

• A novel multi-label learning algorithm was developed, incorporating a voting mechanism from various methods to ensure robust predictions.
• Sequence frequency and position features were extracted to capture characteristic patterns of pure subtypes and CRFs.
• The algorithm was applied to a dataset of 7185 HIV-1 sequences, including 5530 pure subtype and 1655 CRF sequences.

Main Results:

• The developed algorithm achieved high accuracy (up to 99%) in predicting the complete set of subtype labels for HIV-1 recombinant forms.
• The method demonstrated effectiveness in identifying the multiple genetic sources contributing to CRF sequences.
• A small number of incorrect predictions were found to be nearly complete, indicating high confidence in the genuine labels.

Conclusions:

• The multi-label learning approach provides an accurate and reliable method for dissecting the genetic origins of HIV-1 CRFs.
• This tool can significantly aid researchers in understanding HIV-1 evolution and sequence diversity.
• The algorithm shows promise for identifying novel subtypes and tracking the emergence of new genetic lineages.