Supervised fine-tuning enhances unsupervised learning from 45 million amino acids in TCR and peptide sequences

Puntos clave

• This research aims to investigate the effects of supervised fine-tuning on unsupervised learning from a large dataset of amino acid sequences.
• Analyzed 45 million amino acid sequences from TCR and peptide datasets.
• Applied supervised fine-tuning techniques to enhance unsupervised learning models.
• Supervised fine-tuning significantly improved learning efficiency and model accuracy.
• Enhanced models showed better predictive performance compared to traditional unsupervised methods.

Resumen