抗体研究におけるAIタンパク質構造モデリングプラットフォーム

In the rapidly advancing landscape of antibody research, several machine learning techniques have emerged as valuable tools to predict protein-protein interactions. Examples of these methods include creating structural models of proteins, CDR loop structures, and antibody-antigen 3D structure libraries. Artificial intelligence (AI)-driven protein structure modeling platforms such as AlphaFold2 and IgFold are particularly exciting, paving the way to predicting accurate antibody-antigen interactions.

AlphaFold

AlphaFold2 was first showcased by DeepMind in 2020 at CASP14 (14th Critical Assessment of Techniques for Protein Structure Prediction), where AlphaFold2 structures were demonstrated to be vastly more accurate than competing methods by predicting protein folding by constructing in silico models from proteins’ amino acid sequences.¹ It’s median backbone accuracy was 1.84 Å r.m.s.d.95 higher than the next best performing method. AlphaFold2 directly processes multiple sequence alignments for highly accurate protein structure prediction and has been used in applications in molecular replacement and interpreting cryogenic electron microscopy maps.^2,3 

In antibody research, Yin et al. (2022) used AlphaFold2 to model antibody-antigen complexes but had low success rates, finding that AF2 faces difficulties in handling adaptive immune recognition. Gao et al. (2022), however, successfully demonstrated that AF2Complex, an AlphaFold2-based system, could be adapted to predict the structure of protein complexes at much higher accuracy than classical docking approaches.⁴

IgFold

In 2023, Ruffolo et al. developed a fast deep learning method for antibody structure prediction at an even better quality than AlphaFold, and at a much faster pace (under 25 seconds).⁵ By using a pre-trained language model on 558 million natural antibody sequences and graph networks, IgFold directly predicts backbone atomic coordinates which define antibody structure.

This breakthrough allows for potential antibody-antigen prediction accuracy that was previously infeasible. The researchers were able to predict structures for 1.4 million paired antibody sequences, providing structural insights to 500 times more antibodies than those that had been experimentally characterized. In addition, it offers the capabilities for robust incorporation of template structures and support for nanobody modeling. As of now, the main barrier in further developing these AI models is the low availability of experimentally determined structures to help predict how an antibody may bind to a target antigen.

At Biointron, we are dedicated to accelerating your antibody discovery, optimization, and production needs. Our team of experts can provide customized solutions that meet your specific research needs. Contact us to learn more about our services and how we can help accelerate your research and drug development projects.

References:

1. Graves, J., Byerly, J., Priego, E., Makkapati, N., Parish, S. V., Medellin, B., & Berrondo, M. (2020). A Review of Deep Learning Methods for Antibodies. Antibodies, 9(2). https://doi.org/10.3390/antib9020012

2. Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S. A., Ballard, A. J., Cowie, A., Nikolov, S., Jain, R., Adler, J., Back, T., . . . Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583-589. https://doi.org/10.1038/s41586-021-03819-2

3. Yin, R., Feng, B. Y., Varshney, A., & Pierce, B. G. (2022). Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants. Protein Science, 31(8), e4379. https://doi.org/10.1002/pro.4379

4. Gao, M., Nakajima An, D., Parks, J. M., & Skolnick, J. (2022). AF2Complex predicts direct physical interactions in multimeric proteins with deep learning. Nature Communications, 13(1), 1-13. https://doi.org/10.1038/s41467-022-29394-2

5. Ruffolo, J. A., Chu, L., Mahajan, S. P., & Gray, J. J. (2023). Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies. Nature Communications, 14(1), 1-13. https://doi.org/10.1038/s41467-023-38063-x