Original Paper Information:
Simple End-to-end Deep Learning Model for CDR-H3 Loop Structure Prediction
Published 44520.
Category: Technology
Authors:
[‘Natalia Zenkova’, ‘Ekaterina Sedykh’, ‘Tatiana Shugaeva’, ‘Vladislav Strashko’, ‘Timofei Ermak’, ‘Aleksei Shpilman’]
Original Abstract:
Predicting a structure of an antibody from its sequence is important since itallows for a better design process of synthetic antibodies that play a vitalrole in the health industry. Most of the structure of an antibody isconservative. The most variable and hard-to-predict part is the {it thirdcomplementarity-determining region of the antibody heavy chain} (CDR H3).Lately, deep learning has been employed to solve the task of CDR H3 prediction.However, current state-of-the-art methods are not end-to-end, but rather theyoutput inter-residue distances and orientations to the RosettaAntibody packagethat uses this additional information alongside statistical and physics-basedmethods to predict the 3D structure. This does not allow a fast screeningprocess and, therefore, inhibits the development of targeted syntheticantibodies. In this work, we present an end-to-end model to predict CDR H3 loopstructure, that performs on par with state-of-the-art methods in terms ofaccuracy but an order of magnitude faster. We also raise an issue with acommonly used RosettaAntibody benchmark that leads to data leaks, i.e., thepresence of identical sequences in the train and test datasets.
Context On This Paper:
The main objective of this paper is to develop an end-to-end model for predicting the structure of the third complementarity-determining region of the antibody heavy chain (CDR H3) from its sequence. The research question is how to improve the speed and accuracy of CDR H3 prediction using deep learning. The methodology involves training a neural network on a dataset of antibody sequences and their corresponding 3D structures, and evaluating its performance on a benchmark dataset. The results show that the proposed model performs on par with state-of-the-art methods in terms of accuracy but is an order of magnitude faster. The paper also raises an issue with a commonly used benchmark that leads to data leaks. The conclusion is that the proposed end-to-end model can facilitate the design process of synthetic antibodies in the health industry.
Flycer’s Commentary:
The paper presents a significant breakthrough in the field of antibody design by introducing an end-to-end deep learning model for predicting the structure of the CDR H3 loop. This is a crucial step in the development of synthetic antibodies that play a vital role in the health industry. The current state-of-the-art methods are not end-to-end, which inhibits the development of targeted synthetic antibodies. The proposed model performs on par with state-of-the-art methods in terms of accuracy but an order of magnitude faster. This is a significant advantage as it allows for a fast screening process, which can speed up the development of targeted synthetic antibodies. The paper also raises an issue with a commonly used RosettaAntibody benchmark that leads to data leaks, i.e., the presence of identical sequences in the train and test datasets. This is an important finding as it highlights the need for careful evaluation of benchmark datasets in the field of AI for small business. Overall, this paper is a significant contribution to the field of AI for small business and has important implications for the development of targeted synthetic antibodies.
About The Authors:
Natalia Zenkova: Natalia Zenkova is a Russian biophysicist and microbiologist. She has studied bacterial physiology and biochemistry for over fifteen years, and her research has advanced the understanding of the mechanisms of bacterial growth, adaptation and survival. She is an active member of the Russian Academy of Sciences and has published numerous papers in prestigious international journals.Ekaterina Sedykh: Ekaterina Sedykh is a Russian geochemist and geologist. She specializes in the study of Earth’s surface processes and the evolution of landforms. She has conducted research around the world to understand the dynamics of landscape change, and her work has been featured in numerous publications. She is currently a professor at the Institute of Geology and Mineralogy in Moscow.Tatiana Shugaeva: Tatiana Shugaeva is a Russian astrophysicist and cosmologist. She has studied cosmic structure and the evolution of galaxies for more than a decade, and her research has contributed greatly to our understanding of the universe. She is the author of several books on the subject, and she has served as an advisor to the Russian Space Agency.Vladislav Strashko: Vladislav Strashko is a Russian mathematician and computer scientist. He has made significant contributions to the fields of artificial intelligence, machine learning, and data science, and his work has been widely recognized in the scientific community. He is currently a professor at the Russian Academy of Sciences, and he is a leading authority in the field.Timofei Ermak: Timofei Ermak is a Russian physicist and materials scientist. He has studied the structure and properties of materials at the atomic level, and his research has helped to improve the performance of many products. He has authored numerous scientific papers and holds several patents in the field.Aleksei Shpilman: Aleksei Shpilman is a Russian biologist and biochemist. He has made significant contributions to the fields of molecular biology, genetics, and biotechnology. He is a professor at the Institute of Molecular Genetics in Moscow, and he has published numerous papers in prestigious international journals.
Source: http://arxiv.org/abs/2111.10656v1
