Gabriella O. Estevam

Recent advances in protein language models, geometric deep learning, and structure prediction frameworks have the potential to accelerate computational drug discovery. However, it remains unclear how well these methods generalize beyond their training sets. Mutation-dependent drug responses present a challenge to these methods because there are often large effects from relatively small perturbations. Here, we systematically benchmark diverse mutational effect and drug-target affinity prediction models using deep mutational scanning datasets spanning 11 inhibitor conditions for the MET kinase domain. Across both mutational effect prediction and drug-target affinity tasks, current AI models produce largely ligand-invariant predictions despite experimentally distinct inhibitor-dependent fitness landscapes. Ligand-aware geometric models, including LigandMPNN and LASErMPNN, partially recover experimentally validated resistance-associated hotspots, highlighting the importance of explicitly modeling protein-ligand interactions and local structural flexibility. In contrast, deep learning affinity prediction models underestimate mutation-induced changes in binding affinity. This deficiency likely reflects mutation-invariant structural priors derived from co-folded structure prediction frameworks and the absence of mutation-conditioned supervision in current affinity datasets. Finally, we further extended our analysis to several kinase-inhibitor deep mutational scanning datasets and confirm the overall trends. Our findings suggest that a primary limitation in improving performance on this task lies in the scarcity of mutation-conditioned protein-ligand training data rather than model architecture alone.

Receptor Tyrosine Kinases (RTKs) are single-pass transmembrane receptors whose activation is tightly regulated by intra-domain interactions within both their extracellular and intracellular regions. The intracellular juxtamembrane domain, which links the transmembrane and kinase domains, often plays a critical role in modulating kinase activity. The MET receptor, activated by Hepatocyte Growth Factor (HGF), requires precise regulation to support normal development and wound healing, but becomes a potent oncogene when overexpressed or mutated. A common oncogenic lesion in MET, caused by exon 14 skipping, leads to partial deletion of its unusually long intracellular juxtamembrane domain and is frequently detected in non-small cell lung cancer (NSCLC), as well as pancreatic, liver and brain cancers. Despite its length and abundance of post-translational modifications, the functional role of the MET juxtamembrane domain has remained poorly understood. We have uncovered that this segment regulates the kinetics of MET kinase activation. Specifically, we found that a membrane-proximal, N-terminal region of the juxtamembrane domain accelerates activation loop phosphorylation promoting kinase transition to an active state. This regulation does not depend on the oligomeric state of MET but likely acts allosterically to enhance autophosphorylation of the kinase domain. Notably, this function is absent in the closely related MST1R/RON RTK, suggesting it is a unique feature of the MET receptor. Together, these findings uncover a previously unrecognized layer of MET regulation with potential implications for the development of selective therapies targeting MET-driven cancers.

Mutations in the kinase and juxtamembrane domains of the MET Receptor Tyrosine Kinase are responsible for oncogenesis in various cancers and can drive resistance to MET-directed treatments. Determining the most effective inhibitor for each mutational profile is a major challenge for MET-driven cancer treatment in precision medicine. Here, we used a deep mutational scan (DMS) of ~5764 MET kinase domain variants to profile the growth of each mutation against a panel of 11 inhibitors that are reported to target the MET kinase domain. We validate previously identified resistance mutations, pinpoint common resistance sites across type I, type II, and type I ½ inhibitors, unveil unique resistance and sensitizing mutations for each inhibitor, and verify non-cross-resistant sensitivities for type I and type II inhibitor pairs. We augment a protein language model with biophysical and chemical features to improve the predictive performance for inhibitor-treated datasets. Together, our study demonstrates a pooled experimental pipeline for identifying resistance mutations, provides a reference dictionary for mutations that are sensitized to specific therapies, and offers insights for future drug development.

Mutations in the kinase and juxtamembrane domains of the MET Receptor Tyrosine Kinase are responsible for oncogenesis in various cancers and can drive resistance to MET-directed treatments. Determining the most effective inhibitor for each mutational profile is a major challenge for MET-driven cancer treatment in precision medicine. Here, we used a deep mutational scan (DMS) of ∼5,764 MET kinase domain variants to profile the growth of each mutation against a panel of 11 inhibitors that are reported to target the MET kinase domain. We validate previously identified resistance mutations, pinpoint common resistance sites across type I, type II, and type I ½ inhibitors, unveil unique resistance and sensitizing mutations for each inhibitor, and verify non-cross-resistant sensitivities for type I and type II inhibitor pairs. We augment a protein language model with biophysical and chemical features to improve the predictive performance for inhibitor-treated datasets. Together, our study demonstrates a pooled experimental pipeline for identifying resistance mutations, provides a reference dictionary for mutations that are sensitized to specific therapies, and offers insights for future drug development.

CRISPR-Cas enzymes must recognize a protospacer-adjacent motif (PAM) to edit a genomic site, significantly limiting the range of targetable sequences in a genome. Machine learning-based protein engineering provides a powerful solution to efficiently generate Cas protein variants tailored to recognize specific PAMs. Here, we present Protein2PAM, an evolution-informed deep learning model trained on a dataset of over 45,000 CRISPR-Cas PAMs. Protein2PAM rapidly and accurately predicts PAM specificity directly from Cas proteins across Type I, II, and V CRISPR-Cas systems. Using in silico deep mutational scanning, we demonstrate that the model can identify residues critical for PAM recognition in Cas9 without utilizing structural information. As a proof of concept for protein engineering, we employ Protein2PAM to computationally evolve Nme1Cas9, generating variants with broadened PAM recognition and up to a 50-fold increase in PAM cleavage rates compared to the wild-type under in vitro conditions. This work represents the first successful application of machine learning to achieve customization of Cas enzymes for alternate PAM recognition, paving the way for personalized genome editing.

MET is a receptor tyrosine kinase (RTK) responsible for initiating signaling pathways involved in development and wound repair. MET activation relies on ligand binding to the extracellular receptor, which prompts dimerization, intracellular phosphorylation, and recruitment of associated signaling proteins. Mutations, which are predominantly observed clinically in the intracellular juxtamembrane and kinase domains, can disrupt typical MET regulatory mechanisms. Understanding how juxtamembrane variants, such as exon 14 skipping (METΔEx14), and rare kinase domain mutations can increase signaling, often leading to cancer, remains a challenge. Here, we perform a parallel deep mutational scan (DMS) of the MET intracellular kinase domain in two fusion protein backgrounds: wild-type and METΔEx14. Our comparative approach has revealed a critical hydrophobic interaction between a juxtamembrane segment and the kinase ⍺C-helix, pointing to potential differences in regulatory mechanisms between MET and other RTKs. Additionally, we have uncovered a β5 motif that acts as a structural pivot for the kinase domain in MET and other TAM family of kinases. We also describe a number of previously unknown activating mutations, aiding the effort to annotate driver, passenger, and drug resistance mutations in the MET kinase domain.

Mutations in the kinase and juxtamembrane domains of the MET Receptor Tyrosine Kinase are responsible for oncogenesis in various cancers and can drive resistance to MET-directed treatments. Determining the most effective inhibitor for each mutational profile is a major challenge for MET-driven cancer treatment in precision medicine. Here, we used a deep mutational scan (DMS) of ∼5,764 MET kinase domain variants to profile the growth of each mutation against a panel of 11 inhibitors that are reported to target the MET kinase domain. We validate previously identified resistance mutations, pinpoint common resistance sites across type I, type II, and type I ½ inhibitors, unveil unique resistance and sensitizing mutations for each inhibitor, and verify non-cross-resistant sensitivities for type I and type II inhibitor pairs. We augment a protein language model with biophysical and chemical features to improve the predictive performance for inhibitor-treated datasets. Together, our study demonstrates a pooled experimental pipeline for identifying resistance mutations, provides a reference dictionary for mutations that are sensitized to specific therapies, and offers insights for future drug development.

MET is a receptor tyrosine kinase (RTK) responsible for initiating signaling pathways involved in development and wound repair. MET activation relies on ligand binding to the extracellular receptor, which prompts dimerization, intracellular phosphorylation, and recruitment of associated signaling proteins. Mutations, which are predominantly observed clinically in the intracellular juxtamembrane and kinase domains, can disrupt typical MET regulatory mechanisms. Understanding how juxtamembrane variants, such as exon 14 skipping (METΔEx14), and rare kinase domain mutations can increase signaling, often leading to cancer, remains a challenge. Here, we perform a parallel deep mutational scan (DMS) of the MET intracellular kinase domain in two fusion protein backgrounds: wild type and METΔEx14. Our comparative approach has revealed a critical hydrophobic interaction between a juxtamembrane segment and the kinase ⍺C-helix, pointing to potential differences in regulatory mechanisms between MET and other RTKs. Additionally, we have uncovered a β5 motif that acts as a structural pivot for the kinase domain in MET and other TAM family of kinases. We also describe a number of previously unknown activating mutations, aiding the effort to annotate driver, passenger, and drug resistance mutations in the MET kinase domain.

In hereditary papillary renal cell carcinoma (HPRCC), the hepatocyte growth factor receptor (MET) receptor tyrosine kinase (RTK) mutations recorded to date are located in the kinase domain and lead to constitutive MET activation. This contrasts with MET mutations identified in non-small cell lung cancer (NSCLC), which lead to exon 14 skipping and deletion of a regulatory domain: in this latter case, the mutated receptor still requires ligand stimulation. Sequencing of MET in samples from 158 HPRCC and 2808 NSCLC patients revealed ten uncharacterized mutations. Four of these, all found in HPRCC and leading to amino acid substitutions in the N-lobe of the MET kinase, proved able to induce cell transformation, which was further enhanced by hepatocyte growth factor (HGF) stimulation: His1086Leu, Ile1102Thr, Leu1130Ser and Cis1125Gly. Similar to the variant resulting in MET exon 14 skipping, the two N-lobe MET variants His1086Leu and Ile1102Thr were found to require stimulation by HGF in order to strongly activate downstream signaling pathways and epithelial cell motility. The Ile1102Thr mutation also displayed transforming potential, promoting tumor growth in a xenograft model. In addition, the N-lobe-mutated MET variants were found to trigger a common HGF-stimulation-dependent transcriptional program, consistent with an observed increase in cell motility and invasion. Altogether, this functional characterization revealed that N-lobe variants still require ligand stimulation, in contrast to other RTK variants. This suggests that HGF expression in the tumor microenvironment is important for tumor growth. The sensitivity of these variants to MET inhibitors opens the way for use of targeted therapies for patients harboring the corresponding mutations.

Deep mutational scanning (DMS) enables functional insight into protein mutations with multiplexed measurements of thousands of genetic variants in a protein simultaneously. The small sample size of DMS renders classical statistical methods ineffective, for example, p-values cannot be correctly calibrated when treating variants independently. We propose Rosace , a Bayesian framework for analyzing growth-based deep mutational scanning data. Rosace leverages amino acid position information to increase power and control the false discovery rate by sharing information across parameters via shrinkage. To benchmark Rosace against existing methods, we developed Rosette , a simulation framework that simulates the distributional properties of DMS. Further, we show that Rosace is robust to the violation of model assumptions and is more powerful than existing tools under Rosette simulation and real data.