Biology subjects

Al-Lazikani, B.

Publications and source records attributed to Al-Lazikani, B..

8 recordsLinked to original sources

bioRxiv2026

PocketBagger: Generalizable pocket druggability prediction via positive-unlabeled learning

Abstract SummaryReliable structure-based prediction of small-molecule druggability is hindered by a fundamental labeling problem. Experimentally confirmed liganded sites (positives) are observable, but credible "undruggable" pockets (negatives) are almost impossible to define. Standard supervised machine learning consequently relies on arbitrary definitions of undruggable, leading to bias and false negatives. Here we introduce PocketBagger, a positive-unlabeled (PU) learning framework for pocket druggability prediction trained exclusively on experimentally determined Protein Data Bank1 (PDB) structures. PocketBagger uses PU bagging to learn key features associated with reliable druggable pockets and considers all remaining pockets in the structurally characterized proteome as unlabeled. We demonstrate the capability of PocketBagger through the training of a simple Random Forest classifier and demonstrate its power in recall (0.804), even when challenged with increasingly difficult generalizability assessments and entire protein-family hold outs. We benchmark and demonstrate the added value of PU learning by comparing PocketBagger to a leading deep-learning predictor. However, PocketBagger is intended to be used as a framework for any model architecture. Along with the code, the data generated by PocketBagger are deployed in canSAR.ai, providing scalable, generalizable pocket druggability predictions to the drug discovery community.

bioinformatics↗

bioRxiv2025

A Chemoproteomic Atlas of the Human Purine Interactome for Regioselective Ligand Discovery

Purines are essential bioactive molecules that interact with a large fraction of the human proteome. Despite their importance, the scope of actionable purine-binding pockets for ligand discovery remains limited. Here, we developed a quantitative chemoproteomics platform using sulfonyl-purine (SuPUR) chemistry to produce a massive and functional map of the human purine interactome. The SuPUR platform captured 31,000+ targetable tyrosine and lysine sites, representing the most comprehensive beyond cysteine chemoproteomics database for enabling protein ligand discovery. SuPUR ligands that bind through a regioselective fashion serve as enabling starting points for developing potent (nanomolar) and proteome-wide-selective modulators of enzymatic and protein-protein interaction function. Phenotypic screening identified a site-specific (Y237) and regioselective SuPUR ligand of ACAT2 to reveal an unexpected metabolic dependency in cancer cells. A crystal structure of SuPUR ligand-bound ACAT2 revealed the purine group binds deep in the CoA pocket forming key interactions with catalytic residues via a water bridge to guide future structure-based ligand design.

biochemistry↗

bioRxiv2025

Distinct druggable biological processes in early-onset prostate cancer

Despite advances in understanding and treating Prostate Cancer (PCa), there has been little effort to systematically map the biology distinguishing Early-(EOPCa) and Late-(LOPCa) onset PCa. Around 25% of EOPCa cases present with metastatic spread or aggressive disease with earlier metastatic development. Some available lines of therapy are extending treatment trajectories and prolonging lives. However, there remains a critical clinical need to identify new therapeutic targets for EOPCa where life expectancy necessitates safer, more targeted treatment options. To our knowledge, here we present the largest systematic analysis of molecular profiles in EOPCa versus LOPCa, employing machine-learning-enabled algorithms to identify distinguishing biology and druggable targets for each age group. Distinct stromal signatures are uncovered in EOPCa, which are used to propose therapeutic opportunities herein. Moreover, our analysis identifies 50 druggable targets, 11 of which we confirm in PCa cell line genetic/pharmacological perturbation data. These findings provide the first specific, testable hypotheses in EOPCa, offering avenues for experimental validation and potential therapeutic exploitation, and, more generally, shed light on the intricate and distinguished molecular profile of this aggressive, poorly understood disease. One Sentence SummaryMachine learning-enabled algorithms were utilized to identify distinguishing biology and associated druggable targets for early-onset prostate cancers.

cancer biology↗

bioRxiv2024

Probabilistic graph-based model uncovers previously unseen druggable vulnerabilities in major solid cancers

Over half cancer patients lack safe, effective, targeted therapies despite abundant molecular profiling data. Statistically recurrent cancer drivers have provided fertile ground for drug discovery where they exist. But in rare, complex, and heterogeneous cancers, strong driver signals are elusive. Moreover, therapeutically exploitable molecular vulnerabilities extend beyond classical drivers. Here we describe a novel, integrative, generalizable graph-based, cooperativity-led Markov chain model, A3D3as MVP (Adaptive AI-Augmented Drug Discovery and Development Molecular Vulnerability Picker), to identify and prioritize key druggable molecular vulnerabilities in cancer. The algorithm exploits cooperativity of weak signals within a cancer molecular network to enhance the signal of true molecular vulnerabilities. We apply A3D3as MVP to 19 solid cancer types and demonstrate that it outperforms standard approaches for target hypothesis generation by >3-fold as benchmarked against cell line genetic perturbation and drug screening data. Importantly, we demonstrate its ability to identify non-driver druggable vulnerabilities and highlight 43 novel or emergent druggable targets for these tumors.

cancer biology↗

bioRxiv2020

Evolution of kinase polypharmacology across HSP90 drug discovery

Most small molecules interact with several target proteins but this polypharmacology is seldom comprehensively investigated or explicitly exploited during drug discovery. Here, we use computational and experimental methods to systematically characterize the kinase cross-pharmacology of representative HSP90 inhibitors. We demonstrate that the resorcinol clinical candidates ganetespib and, to a lesser extent, luminespib, display unique off-target kinase pharmacology as compared to other HSP90 inhibitors. We also demonstrate that polypharmacology evolved during the optimisation to discover luminespib and that the hit, leads and clinical candidate all have different polypharmacological profiles. We conclude that the submicromolar target inhibition of protein kinases by ganetespib may have potential clinical significance and we recommend the computational and experimental characterization of polypharmacology earlier in drug discovery projects to unlock new multi-target drug design opportunities.

pharmacology and toxicology↗

bioRxiv2019

Signalling involving MET and FAK supports cell division independent of the activity of the cell cycle-regulating CDK4/6 kinases

Deregulation of the cyclin-dependent kinases 4 and 6 (CDK4/6) is highly prevalent in cancer yet inhibitors against these kinases currently show use in restricted tumour contexts. The extent to which cancers depend on CDK4/6 and what may undermine such dependency is poorly understood. Here we report that signalling engaging the MET proto-oncogene receptor tyrosine kinase/focal adhesion kinase (FAK) axis leads to CDK4/6-independent CDK2-activation, involving as a critical mechanistic events loss of the CDK inhibitor p21CIP1 and gain of its regulator, the ubiquitin ligase subunit SKP2. Combined inhibition of MET/FAK and CDK4/6 eliminates proliferation capacity of cancer cells in culture, and enhances tumour growth inhibition in vivo. Activation of the MET/FAK axis is known to arise through cancer extrinsic and intrinsic cues. Our work predicts that such cues support cell division independent of the activity of the cell cycle-regulating CDK4/6 kinases and identifies MET/FAK as a tractable route to broaden the utility of CDK4/6 inhibitor-based therapies in the clinic.

cancer biology↗

bioRxiv2019

The off-target kinase landscape of clinical PARP inhibitors

Four PARP inhibitors have been approved by the FDA as cancer therapeutics, yet there are several clinical settings where no strong rationale exists to inform clinicians on which to choose in terms of clinical effectiveness and toxicity. The four drugs have largely similar PARP family inhibition profiles, but several differences at the molecular and clinical level have been reported that remain poorly understood. We previously hypothesized that PARP inhibitors could have an inherent capacity to inhibit kinases off-target. Here, we characterise the off-target kinase landscape of four FDA-approved PARP drugs. We demonstrate that all four PARP inhibitors have a unique polypharmacological profile across the kinome. Niraparib and rucaparib inhibit DYRK1s, CDK16 and PIM3 at clinically achievable submicromolar concentrations. These represent the most potently inhibited off-targets of PARP inhibitors identified to date and should be investigated further to clarify their potential implications for efficacy and safety in the clinic, including use of PARP inhibitors in combination with other agents, including immunotherapy.

pharmacology and toxicology↗

bioRxiv2017

Objective, Quantitative, Data-Driven Assessment of Chemical Probes

Chemical probes are essential tools for understanding biological systems and for target validation, yet selecting tools for biomedical research is largely biased and subjective. Here we describe the Probe Miner: Chemical Probes Objective Assessment resource - capitalising on the plethora of public medicinal chemistry data to empower quantitative, objective, Big Data-driven assessment of chemical probes. We assess >1.8m compounds for their suitability as chemical tools against 2,220 human targets and dissect their biases and limitations.

bioinformatics↗