Specific and intrinsic sequence patterns extracted by deep learning from intra-protein binding and non-binding...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• The DNA double helix model relies on specific nucleotide pairing rules (A-T, C-G), forming the basis of the genetic code.
• No analogous specific binding rules have been identified for protein peptide fragments.

Purpose of the Study:

• To investigate the existence of intrinsic sequence patterns within intra-protein binding peptide fragments.
• To develop a method for extracting these patterns and compare them to DNA binding rules.
• To explore the implications of these patterns for protein folding and interactions.

Main Methods:

• Utilized a deep learning algorithm to extract intrinsic sequence patterns from intra-protein binding peptide fragments.
• Classified millions of binding and non-binding peptide fragments derived from protein X-ray structures.

Main Results:

• Identified distinct, intrinsic sequence patterns specific to intra-protein binding peptide fragments, differentiating them from non-binding fragments.
• Achieved up to 93% accuracy in classifying binding and non-binding peptide fragments using the deep learning model.
• Observed a resemblance between these protein fragment patterns and the DNA double helix model.

Conclusions:

• Specific binding rules for protein peptide fragments have been discovered, analogous to DNA base pairing.
• These intrinsic sequence patterns may drive protein folding and protein-protein interactions.
• The findings hold significant potential for the design and development of peptide, protein, and antibody-based therapeutics.