Artificial intelligence approaches for design/discovery of therapeutic proteins: a systematic review

Background and aims: The development of novel proteins for disease diagnosis, alleviation, andimproved health attributes has been enabled by protein engineering. Artificial intelligence (AI)has significant potential to improve smart protein engineering without a complete understandingof molecular mechanisms.Here, we present a review of recent AI techniques and approaches for modifying proteins at differentlevels, which find therapeutically relevant applications.Methods: A systematic review was carried out by two reviewers independently and manuallysearching English databases (PubMed, Bireme, EBSCO, OVID, Scopus, and Web of Science)for data till March ۲۰۲۳, resulting in ۱۶۶ research articles. After the quality screening, a finalselection of ۱۳۳ articles was made for further analysis. Databases were searched using the terms‘artificial intelligence’, ‘cancer’, ‘ therapeutic protein’, ‘ design of protein’, ‘protein design andengineering’, ‘machine learning’, ‘protein prediction’, and ‘drug design’.Results: As a result of our study, we identified the following outcomes: AI applications are mainlydivided into four categories: ۱) genomics, ۲) protein structure and function, ۳) protein designand evolution, and ۴) drug design. Among the ML algorithms and databases used, the most commonmethod was supervised learning (۸۵%). The most common databases used for ML modelswere PDB and UniProtKB/Swiss-Prot (۲۱ and ۸%, respectively). Finally, we describe the currentapplications of AI-assisted protein engineering, as well as the prospects of this field in the future.Our main finding is that, as of today, there are no research road maps serving as guides to addressgaps in our knowledge of the AI–PS binomial. A discussion of current limitations and methodsis presented, along with a look at future directions. Across a range of disease prevalences, AIsystems can deliver the main benefits of biopsy avoidance while maintaining high specificities.Conclusion: Overall, AI is a valuable tool for the screening and designing of therapeutic proteins,with particular attention to anticancer proteins and cutting-edge AI technology embedded, leadingthe progress of innovative therapeutics for challenging diseases.

کلیدواژه ها:

artificial intelligence ، therapeutic protein ، protein engineering ، protein design and engineering ، machine learning ، deep learning