Authors: Nayoung Kim, Minsu Kim, Jinkyoo Park

Published on: February 08, 2024

Impact Score: 8.3

Arxiv code: Arxiv:2402.05982

Summary

• What is new: Introduces a new graphical model, AGN, that integrates a pre-trained protein language model for antibody design, significantly outperforming previous methods.
• Why this is important: Existing deep learning methods in antibody design have limitations in utilizing general protein knowledge and assume a graphical model that contradicts empirical protein findings.
• What the research proposes: AGN employs a two-step sequence generation and structure prediction process, incorporating a novel composition-based regularization to reduce token repetition.
• Results: AGN outperforms state-of-the-art results in benchmark experiments and establishes a Pareto frontier over current methods.

Technical Details

Technological frameworks used: AGN (Anfinsen Goes Neural)

Models used: Pre-trained protein language model (pLM) and Graph Neural Network (GNN)

Data used: nan

Potential Impact

Pharmaceutical and biotech companies involved in therapeutic antibody design could be significantly impacted, improving efficiency and effectiveness in developing new treatments.

Want to implement this idea in a business?

We have generated a startup concept here: ProGen Therapeutics.