Research contributions and leadership

From spectra to systems biology

My doctoral work at CNIC began with a practical problem: quantitative proteomics does not produce a flat table. Several spectra can support a peptide, several peptides a protein, and several proteins a complex or functional category. Each level carries uncertainty from the one below it. I developed a statistical framework that integrated those levels while propagating the associated error, providing the foundation for the methods and software that followed.

I then co-developed the Systems Biology Triangle (SBT), a model designed natively for pairwise quantitative-proteomics experiments. Instead of transferring a pathway method from another omics setting, SBT works with the hierarchical structure of mass-spectrometry data to detect coordinated changes at protein and functional-category levels.

The third step was to make the approach usable beyond one analysis. I created the original SanXoT package and led its development as modular research software. Its configurable workflows joined protein quantification, systems-biology analysis, post-translational modifications, and integration of technical or biological replicates. The statistical framework, SBT and SanXoT therefore form one connected programme: from a model of uncertainty, through biological interpretation, to software that other researchers could run and adapt.

Public record: statistical framework, SBT model, SanXoT source code, and CNIC documentation.

Open software that keeps evolving

I prioritise open access and open-source code because sharing makes a computational method more useful: researchers can inspect it, reproduce it and build on it. Open software also creates the conditions for synergy, allowing individual contributions to become part of a larger and continuing body of work. This is why my main, finished programs are available through GitHub rather than existing only as descriptions in papers.

SanXoT provides a particularly clear example of what that choice can enable. After my direct involvement at CNIC ended, a community of researchers continued developing the same methodological environment. The original integrations were carried into iSanXoT, where SBT remains available as a dedicated module and as part of compound WSPP–SBT workflows connecting scans, peptides, proteins or genes, and functional categories. Public documentation, source code and a continuing release history show that the framework remains operational rather than merely archived.

The surrounding CNIC-Proteomics group now contains a broader family of database, workflow and post-translational-modification tools, including bioDataHub and PTM-compass. They are not all extensions of SanXoT itself, but together they reveal an active developer community around the analytical universe in which SanXoT and SBT were created.

Contributing to Comet-PTM

In parallel, I contributed to the creation of Comet-PTM, a branch of the open-source Comet search engine, from the SEQUEST lineage, for identifying and locating post-translational modifications found through open-search strategies. The work supported a Cell Reports study of the modified proteome. Parts of the code I contributed were later incorporated into the main Comet codebase by its author, Jimmy Eng. That movement from a specialist branch into upstream software is another route by which a contribution can continue independently of its original project.

Public record: iSanXoT documentation and workflows, iSanXoT release history, CNIC-Proteomics repositories, Comet-PTM source code, and the associated Cell Reports study.

Chromatinsight: machine learning for comparative epigenomics

After finishing my PhD, I moved in 2018 to Newcastle University, looking for new horizons in computational biology. For me, it was bigger the change of moving from proteomics to epigenomics than moving between countries. It meant learning a new data landscape: the basic unit, the starting point, which for me had been mass spectra, became now sequences, a very different way of thinking!

However, the underlying question remained: how can a complex molecular profile be turned into an interpretable comparison? The transition gave me an opportunity to apply the same creative approach (building methods around the structure of the biological problem) in a different field. During my postdoctoral work with Daniel Rico, I developed Chromatinsight to identify epigenomic features that distinguish two groups of samples. The model is based on a random forest algorithm (although this can be replaced by other methods) to learn from chromatin-state segmentations produced with ChromHMM, allowing combinations of marks and genomic contexts to be prioritised without reducing the analysis to one signal at a time.

I built the initial Python implementation and made it available with a tutorial, test data and an open-source licence. A companion R toolkit supports the preparation and exploration of Chromatinsight analyses. The first biological application compared male and female epigenomic profiles, but the design is general: the two groups can represent different conditions, phenotypes or experimental contexts.

Chromatinsight has also continued beyond the Newcastle phase in which I initiated it. It is currently under further development at CABIMER within Daniel Rico's Computational Epigenomics and Cell Identity group. Its continued development by that group, together with the public code and reusable supporting tools, gives the project a life beyond my own appointment and provides a concrete line of continuity for this contribution.

Public record: Chromatinsight model, tutorial and test data, companion R tools, and the CABIMER research group where the work is being developed further.

The Onsager prominence model

Discovery in my current methodological work begins with a different question: when a treatment changes a disease-associated transcriptomic profile, which biological processes are genuinely restored, which remain altered, and which move further away from the reference state? I am developing the Onsager prominence model to distinguish those response patterns rather than compressing them into a single enrichment result.

The work is being developed within Gavin Richardson's experimental research environment, linking a mathematical framework with treatment-response questions in ageing and cardiovascular biology. While the model is still under development, its public implementation already includes an interactive viewer, an online analysis runner, demonstration data and an explanatory infographic. This makes the method inspectable and testable as it develops.

Explore the work: have a look at the Onsager viewer and analysis runner, or start with the Onsager prominence model infographic (PDF) for a visual introduction. You can also explore the wider current lab context.

Competitive funding and grant contributions

Funding has supported different forms of responsibility across my career: an individual predoctoral fellowship, named co-investigator roles on charity grants, an internally supported project that I originated, and research roles within larger funded programmes. The distinctions matter, so the record below states my role rather than treating every award as equivalent.

Teaching and sharing computational practice

I have contributed to bioinformatics and computational-biology teaching since 2009, beginning with proteomics lectures at the Universidad Francisco de Vitoria and the Universidad Autónoma de Madrid and later teaching undergraduate and postgraduate students at Newcastle University. What connects these activities is an emphasis on making computational analysis usable by researchers who are not bioinformatics specialists.

In 2026, I delivered three lectures on epigenomics and single-cell RNA sequencing for ELTE's practice-oriented Analysis of Omics Data PR summer course in Budapest. I also contributed 12 hours of lectures to the CHARM-EU transnational version of the programme.

The course materials are maintained in a public teaching repository, extending their value beyond the students present in the room. Keeping them open allows learning to continue after the course and turns teaching into a practical form of sharing.