Computational Biology & AI

From raw data to a validated biological signal

Our computational biology team handles the dry-lab side of every project, from bioinformatics pipelines and statistical analysis to AI-driven models designed to uncover the signals that matter in your data. Every result is interpreted by our scientists and connected back to the biology—never delivered as a black box.

What the computational team covers

Bioinformatics pipelines

Alignment, quantification and QC, version-controlled end to end.

  • RNA-seq, WGS/WES, Metagenomics, ChIP-seq, ATAC-seq, Proteomics & Multi-omics Data Analysis
  • Variant calling, quantification & QC reporting
  • Reproducible, containerized runs from raw reads to results

Multi-omics integration

Combining genomics, transcriptomics and other layers into one coherent model.

  • Joint modeling across genomics, transcriptomics & proteomics and metagenomics
  • Batch correction and cross-platform harmonization
  • Multi-layer correlation & network analysis

AI biomarker discovery

Inference models to prioritize candidate biomarkers from complex datasets.

  • Feature selection on high-dimensional omics data
  • Cross-validated classifiers for diagnostic/prognostic signals
  • Robustness testing across independent cohorts

Analysis by research area

How this computational work applies in each sector.

Have a dataset that needs interpreting?

Let's talk about your data and your biological question.

Write to us

Questions we get asked

Working from your data, what you receive, and how a signal is held to be real.

Can you work on data we generated elsewhere?

Yes, and it is a large part of what the computational team does. We start from your raw or processed files, re-run the quality control so that the starting point is documented, and tell you what the data can and cannot support before proposing an analysis.

What do we receive at the end?

A report that answers the biological question, the processed data and result tables, and the figures. The pipeline versions and reference genome are stated so the analysis can be repeated. Anything beyond that — code, notebooks, an interactive dashboard — is agreed in the study documentation rather than assumed.

How do you avoid reporting a signal that does not hold up?

By deciding the analysis plan before looking at the outcome, and by separating discovery from validation. A signal found on one set is tested on data that did not contribute to finding it. Where the cohort is too small for that, we say so and report the finding as exploratory — which is a result, not a failure.

Which reference genome and pipeline versions do you use?

They are fixed per study and written down, because two analyses run on different annotation versions are not comparable and the difference does not show in the report. We state the reference build, the annotation release and the pipeline version with the results.