TFIDF-Random Forest: Prediction of Aptamer-Protein Interacting Pairs

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• Aptamers are short oligonucleotides or peptides with high specificity for molecular targets.
• Their ability to bind proteins is valuable for drug development and controlled therapeutic release.
• Current in vitro and in silico methods for aptamer selection are complex and computationally expensive.

Purpose of the Study:

• To introduce a natural language processing (NLP) approach for data-driven computational aptamer selection.
• To compare the efficiency of the NLP method against existing sequential models with embedding layers.
• To advance the development of more efficient computational methods for aptamer discovery.

Main Methods:

• Transformed DNA/RNA and protein sequences into text format using a sliding window approach.
• Developed and applied a natural language processing model for aptamer-protein binding prediction.
• Compared the performance of the NLP model against a literature-based sequential model.

Main Results:

• The NLP methodology demonstrated notably higher efficiency compared to existing literature methods.
• The data-driven computational selection approach showed marked improvement over previous models.
• Preliminary results indicate significant advancements in computational aptamer selection.

Conclusions:

• The developed NLP approach offers a more efficient and data-driven method for aptamer selection.
• This advancement brings computational methods closer to practical application in drug development.
• The study highlights the potential of NLP in accelerating the discovery of novel aptamers.