Investigators affiliated with the Icahn School of Medicine at Mount Sinai have pinpointed a previously undetected, druggable locus within a protein implicated in oncogenesis. This discovery holds the potential to pave the way for a new cohort of highly targeted therapeutic agents for cancer treatment. Furthermore, this finding sheds light on critical constraints inherent in contemporary artificial intelligence methodologies employed in pharmaceutical research and development.
The research, disseminated in the June 2nd online edition of the Journal of the American Chemical Society [10.1021/jacs.6c05178], centered its investigation on PKMYT1, a protein classification known as a kinase, which plays a crucial role in the regulation of cellular proliferation and division. Given that dysregulation of this process is a hallmark of cancerous conditions, PKMYT1 has emerged as a significant prospective target for novel oncological therapeutics.
The prevailing strategy for developing experimental drugs aimed at inhibiting kinases involves targeting a specific region known as the ATP-binding site – the domain of the protein responsible for utilizing the cellular energetic reserves to execute its function. However, a substantial number of kinases exhibit nearly indistinguishable ATP-binding sites, thereby posing a challenge for therapeutic agents to differentiate effectively between the intended target and other kinases. This lack of specificity can precipitate undesirable adverse effects.
Through a synergistic application of AI-driven protein structural prediction tools and rigorous laboratory experimentation, the research team successfully identified a novel, “occult” pocket within PKMYT1 capable of accommodating molecular binding. This particular site remained undetected by current, cutting-edge AI systems.
Our investigation underscores both the formidable capabilities and the inherent limitations of artificial intelligence in the realm of drug discovery. While AI demonstrated exceptional proficiency in predicting known protein configurations, it failed to identify a completely unanticipated binding cavity that we could only elucidate through empirical methods. This concealed locus may ultimately provide an innovative avenue for the design of more selective anticancer agents.”
Avner Schlessinger, PhD, co-senior and co-corresponding author, Professor of Pharmacological Sciences, Director of the AI Small Molecule Drug Discovery Center, and Associate Director, Mount Sinai Center for Therapeutics Discovery, Icahn School of Medicine at Mount Sinai
These findings imply that proteins such as PKMYT1 possess a greater degree of conformational plasticity than was heretofore understood, undergoing continuous transitions between diverse structural states rather than maintaining a singular, static form. The study additionally indicated that even minute alterations to a molecule’s chemical composition could profoundly influence the manner and location of its interaction with the protein, as reported by the investigators.
The scientific cadre employed the AlphaFold2 AI system for the prediction of potential PKMYT1 structural conformations, subsequently executing virtual screening to identify molecules that might engage with it. Subsequent investigations involved X-ray crystallography, biochemical assays, and cellular analyses to validate the behavior of these molecules across various experimental paradigms.
Additional AI platforms, encompassing AlphaFold3 and Boltz-2, in conjunction with molecular dynamics simulations, were then utilized to ascertain whether contemporary computational methodologies could accurately forecast the newly identified binding modality.
“One of the most unexpected revelations was that a minuscule chemical modification induced a shift in the molecule’s binding behavior, transitioning from interaction within this concealed pocket to engaging with a far more conventional site,” stated co-senior and co-corresponding author Michael Lazarus, PhD, Associate Professor of Pharmacological Sciences, and Associate Director of the Mount Sinai Center for Therapeutics Discovery, at the Icahn School of Medicine at Mount Sinai. “This observation suggests that these proteins exhibit remarkable dynamism and are highly responsive to subtle molecular variations. It further reinforces the indispensable role of experimental validation, even within the current era of AI ubiquity.”
The research principals posit that this work could ultimately contribute to the advancement of more selective therapeutic agents, thereby mitigating some of the toxicity and specificity challenges associated with conventional kinase inhibitors. The research outcomes may also serve to enhance future AI systems by imbuing them with a greater capacity to recognize concealed and dynamic protein conformations that are presently overlooked.
While further rigorous investigation is warranted, these findings establish a significant preliminary framework for the development of prospective treatments targeting this recently discovered locus. The molecular entities identified during the study represent promising foundational candidates for subsequent refinement and evaluation in relevant disease models.
The team’s forthcoming endeavors will involve the synthesis of more potent compounds designed to interact with the newly identified site and an exploration into the potential existence of analogous concealed pockets within other protein kinases associated with cancer. Additionally, they aim to refine computational algorithms to enable AI systems to more accurately predict these elusive protein structures in the future.
Herrington, N. B., et al. (2026). Allosteric Inhibition of PKMYT1 Induces a Unique, Inactive ATP Binding Site Conformation. Journal of the American Chemical Society. DOI: 10.1021/jacs.6c05178. https://pubs.acs.org/doi/10.1021/jacs.6c05178
