In a groundbreaking convergence of bioengineering, evolutionary biology, and computational biology, researchers at The University of Texas at Dallas have unveiled a sophisticated machine-learning model that predicts the behavior of proteases with exceptional accuracy. Proteases, a class of enzymes functioning as molecular scissors, selectively cleave proteins and hold transformative potential for targeted therapies against viruses and cancer. However, designing proteases optimized for specific tasks has long been hampered by unpredictable enzymatic activity. This latest advancement could redefine protein engineering, expediting the development of highly efficient, tailor-made enzymes that surpass current standards.
The model, termed ProSSpeC (Protease Substrate Specificity Calculator), leverages vast evolutionary datasets to analyze how proteases have diversified and adapted over millions of years. By interpreting patterns of amino acid changes across thousands of related enzymes within the Potyviridae family of plant viruses, ProSSpeC pinpoints the critical residues that underlie enzymatic function. This co-evolutionary approach enables the model to assess how alterations in the protease sequence affect substrate recognition and cleavage efficiency without exhaustive laboratory trials.
Traditionally, the engineering of proteases has relied heavily on iterative experimental methods, often entailing laborious trial-and-error screening of numerous variants to identify effective candidates. This process slows therapeutic innovation and demands significant resources. ProSSpeC disrupts this paradigm by computationally forecasting proteolytic activity and specificity, thereby narrowing the selection of promising variants prior to physical synthesis. Such predictive power marks an important step toward rational enzyme design rooted in evolutionary biology.
Beyond its conceptual novelty, ProSSpeC has demonstrated practical superiority. The researchers synthesized several proteases suggested by the model and tested their performance in cellular environments. Remarkably, these engineered enzymes outperformed the widely used tobacco etch virus protease, a common tool in protein purification and pharmaceutical manufacturing. This experimental validation confirms ProSSpeC’s ability not only to predict but to guide the creation of highly functional proteases tailored for biomedical applications.
Central to the success of this work is the interdisciplinary collaboration fostering the integration of computational insights and experimental expertise. Dr. P.C. Dave Dingal, assistant professor of bioengineering, and Dr. Faruck Morcos, associate professor of biological sciences, co-led the study, bringing together domains spanning machine learning, evolutionary modeling, and synthetic biology. Their combined efforts underline how embracing nature’s evolutionary record can reveal design principles obscured in isolated datasets.
The Potyviridae virus family provided an ideal model system for this study due to its extensive diversification and well-characterized proteases. By comparing thousands of variants within this group, ProSSpeC constructs a nuanced map of sequence-function relationships. This map serves as a blueprint to engineer proteases with bespoke substrate preferences and catalytic efficiencies—attributes critical for crafting therapeutic enzymes able to delicately modulate protein interactions implicated in disease.
Biomedical engineering doctoral student Medel B. Lim Suan Jr., a key contributor to the project, highlighted the unique educational opportunity afforded by this project’s dual computational and experimental framework. Participating students gained hands-on experience bridging data-driven modeling and biological assays, illustrating how interdisciplinary research catalyzes innovation and skill development simultaneously.
The implications of this technology are profound. Engineered proteases hold promise not only as components of antiviral or anticancer therapies but also as precision tools in molecular biology, capable of cleaving proteins at exact locations to manipulate protein function or facilitate biotechnology workflows. ProSSpeC’s capacity to accelerate the design of such enzymes could transform multiple fields reliant on protein engineering.
Moreover, incorporating evolutionary constraints into enzyme design leverages natural selection’s empirical wisdom, allowing researchers to navigate the vast sequence space more intelligently. This evolutionary-guided strategy enhances the likelihood that engineered proteases will be both effective and stable within cellular systems, addressing frequent challenges in protein therapeutics development.
The team has secured provisional patents for several of the novel proteases identified, underscoring the translational potential of their findings for industry and medicine. Funding for this research was provided by the National Institutes of Health, the National Science Foundation, and institutional grants from UT Dallas, reflecting the high scientific and societal value placed on this innovative approach.
Published in the prestigious journal Nature Communications on February 26, 2026, this study represents a paradigm shift from empirical enzyme engineering to a data-informed, predictive science. By demonstrating that evolutionary and computational methods can pinpoint protease variants outperforming standard tools, the research heralds a new era of rapid, rational development in protein therapeutics.
In summary, ProSSpeC exemplifies how harnessing evolutionary history through machine learning can unravel complex biomolecular functions and inspire the rational engineering of enzymes with superior capabilities. This advance not only promises to accelerate therapeutic development but also reinforces the transformative power of interdisciplinary science in addressing pressing biomedical challenges.
Subject of Research: Cells
Article Title: Identification and engineering of highly functional potyviral proteases in cells using co-evolutionary models
News Publication Date: 26-Feb-2026
Web References:
ProSSpeC Model: https://coevolutionary.org/prosspec/
Published Article: https://www.nature.com/articles/s41467-026-69961-5
References:
Dingal, P.C.D., Morcos, F., et al. Nature Communications, 2026.
Image Credits: The University of Texas at Dallas
Keywords
Protease, Enzymes, Biomolecular Structure, Biomolecules, Protein Functions, Cell Biology, Evolutionary Biology, Developmental Biology, Computational Biology, Proteins
Tags: accelerating protein therapeutic innovationco-evolutionary analysis of enzymescomputational biology in enzyme designdesigning proteases for cancer treatmentevolutionary biology of proteasesmachine learning in protease predictionPotyviridae family enzyme studypredictive models for enzyme activityprotease substrate specificity calculatorprotein engineering techniquestargeted enzyme therapy developmentviral protease engineering
