Omics Biomarker Discovery and Custom Explainable Modelling for Precision Medicine

High-dimensional omics datasets contain valuable biomarker signals, but converting them into compact, reproducible and clinically meaningful signatures is difficult. I offer longer-term scientific collaborations to biotechnology, pharmaceutical, diagnostic and data-driven health companies seeking to design, execute or strengthen biomarker-development programmes.

My expertise combines bioinformatics, machine learning and survival analysis using transcriptomic, methylation, protein-expression and standard clinical data. I focus on defining clinically relevant endpoints; curating and harmonising multi-cohort datasets; discovering compact biomarker signatures; integrating molecular and clinicopathological variables; building interpretable predictors; and designing rigorous internal and independent validation.

I develop both project-specific computational methods and applications of established statistical and machine-learning techniques. My preferred programming environment is Python. Depending on project needs, I can produce reproducible analysis pipelines, ad hoc research software and documented computational prototypes, as well as advise company teams on methodology and study design.

My research experience includes:

* Breast cancer: explainable prediction of five-year survival from gene-expression profiles, with treatment-specific analyses using the approximately 2,000-patient METABRIC cohort and assessment in external cohorts totalling 601 patients.

* Localised prostate cancer: integration of clinical, transcriptomic, methylation and protein-expression data to develop prognostic fingerprints for five-year disease-free survival.

* Leukaemia: ongoing confidential research on omics-based survival prediction in paediatric acute myeloid leukaemia. Unpublished methodological and biomarker details are not disclosed in this profile.

I can contribute through strategic scientific advice, hands-on computational analysis and independent technical review, including:

* formulation of the clinical question, endpoint and statistical analysis plan;

* cohort selection, data-quality assessment and multi-dataset harmonisation;

* development of custom computational methods and Python software;

* feature selection, biomarker discovery and predictive-model development;

* survival, fixed-horizon and subgroup analyses;

* prevention of data leakage, overfitting and optimistic performance estimates;

* evaluation of discrimination, calibration, robustness and clinical relevance;

* independent-cohort validation and cross-platform reproducibility assessment;

* biological interpretation and evidence review for selected biomarkers;

* technical due diligence or independent review of existing biomarker pipelines;

* preparation of results for publication, partnering and further product development.

My approach gives particular attention to explainability, small biomarker panels, reproducibility across cohorts and transparent communication of limitations. The objective is not merely to optimise performance in one dataset, but to generate evidence that can survive independent testing and support subsequent assay development and clinical evaluation.

I am primarily seeking substantial, longer-term biomarker-development collaborations, preferably structured through an agreement with CNR. I can work alongside company scientists, clinicians, wet-lab specialists and regulatory teams as a scientific adviser, hands-on computational partner or independent reviewer.

I welcome meetings with organisations developing biomarkers in oncology or other diseases and seeking expertise in omics study design, custom computational methodology, Python analysis, prognostic modelling, multi-cohort validation or independent project assessment.