Biology subjects

Mann, G. J.

Publications and source records attributed to Mann, G. J..

4 recordsLinked to original sources

bioRxiv2025

A blueprint for mutation-defined hallmark vulnerabilities across human cancers

Hallmark gene mutations shape cancer cell vulnerabilities and inform drug discovery1-3. A systematic map of hallmark gene mutation-defined cancer dependencies and therapeutic responses is essential to uncover novel targets and refine therapeutic strategies. Here, we present the first pan-cancer blueprint of hallmark vulnerabilities, systematically linking hallmark gene mutation markers to cancer cell dependencies and drug sensitivities across 22 cancer cohorts. We integrated multi-omics data from patient tumours with large-scale CRISPR-Cas9 screens and pharmacologic profiling of over a thousand cancer cell lines. Our analysis revealed the cancer type-specific nature of hallmark gene expression programs, uncovered previously unrecognised mutation-target gene dependencies, and highlighted metabolic programs as a dominant class of functional vulnerabilities. Notably, we identified oxidative phosphorylation (OXPHOS) addiction in CDKN2A-loss lung squamous cell carcinoma (LUSC) and experimentally validated this dependency. Our validation highlights the greater selectivity of CDKN2A-loss LUSC cells to metformin, an FDA-approved antidiabetic drug known for its OXPHOS inhibitory activity. Proteogenomic integration further prioritised targets overexpressed in mutant tumours, constituting therapeutic windows. Pharmacologic profiling identified both oncology and non-oncology agents with selective activity in mutation-defined subgroups, revealing opportunities for drug repurposing. Our machine learning framework, Comet-X, for the first time fully leveraged gene mutation combinations to predict these target dependencies and drug responses. The resulting pan-cancer mutation-dependency map provides a comprehensive resource of hallmark gene targets and candidate therapeutics, stratified by mutation markers, to pave the way for drug development, clinical trial design and discovery research.

cancer biology↗

bioRxiv2024

Federated deep learning enables cancer subtyping by proteomics

Artificial intelligence applications in biomedicine face major challenges from data privacy requirements. To address this issue for clinically annotated tissue proteomic data, we developed a Federated Deep Learning (FDL) approach (ProCanFDL), training local models on simulated sites containing data from a pan-cancer cohort (n=1,260) and 29 cohorts held behind private firewalls (n=6,265), representing 19,930 replicate data-independent acquisition mass spectrometry (DIA-MS) runs. Local parameter updates were aggregated to build the global model, achieving a 43% performance gain on the hold-out test set (n=625) in 14 cancer subtyping tasks compared to local models, and matching centralized model performance. The approachs generalizability was demonstrated by retraining the global model with data from two external DIA-MS cohorts (n=55) and eight acquired by tandem mass tag (TMT) proteomics (n=832). ProCanFDL presents a solution for internationally collaborative machine learning initiatives using proteomic data, e.g., for discovering predictive biomarkers or treatment targets, while maintaining data privacy. Statement of SignificanceA federated deep learning approach applied to human proteomic data, acquired using two distinct proteomic technologies from 40 tumor cohorts from eight countries, enabled accurate cancer histopathological subtyping while preserving data privacy. This approach will enable privacy-compliant development of large-scale proteomic AI models, including foundation models, across institutions globally.

cancer biology↗

bioRxiv2020

Cross-Platform Omics Prediction procedure enables precision medicine in patients with stage-III melanoma

There is no consensus methodology that can account for the variation in omics signatures when they are acquired across different platforms and times. This poses a significant barrier to the implementation of valuable biomarkers into clinical practice. We present a novel procedure (Cross-Platform Omics Prediction) that accounts for these variations and demonstrate its utility in three risk models for different diseases that is suitable for prospective and multi-centre clinical implementation.

bioinformatics↗

bioRxiv2020

Anatomic position determines oncogenic specificity in melanoma

Oncogenic alterations to DNA are not transforming in all cellular contexts1,2. This may be due to pre-existing transcriptional programs in the cell of origin. Here, we define anatomic position as a major determinant of why cells respond to specific oncogenes. Cutaneous melanoma arises throughout the body, whereas the acral subtype arises on the palms of the hands, soles of the feet, or under the nails3. We sequenced the DNA of cutaneous and acral melanomas from a large cohort of human patients and found a specific enrichment for BRAF mutations in cutaneous melanoma but CRKL amplifications in acral melanoma. We modeled these changes in transgenic zebrafish models and found that CRKL-driven tumors predominantly formed in the fins of the fish. The fins are the evolutionary precursors to tetrapod limbs, indicating that melanocytes in these acral locations may be uniquely susceptible to CRKL. RNA profiling of these fin/limb melanocytes, compared to body melanocytes, revealed a positional identity gene program typified by posterior HOX13 genes. This positional gene program synergized with CRKL to drive tumors at acral sites. Abrogation of this CRKL-driven program eliminated the anatomic specificity of acral melanoma. These data suggest that the anatomic position of the cell of origin endows it with a unique transcriptional state that makes it susceptible to only certain oncogenic insults.

cancer biology↗