When a test helps decide treatment, the software matters
A companion diagnostic identifies the patients for whom a particular treatment is appropriate. That makes the analysis behind it clinically consequential and regulatorily in scope — which changes what the bioinformatics has to satisfy.
How this differs from research analysis
The methods are often familiar. The obligations around them are not.
Analytical validation
Sensitivity, specificity, precision, and limit of detection established for the specific variants or markers the test claims, on the specimen types it will actually see.
A defined intended use
The test is validated for a stated population, specimen type, and therapeutic question. Results outside that scope are not supported by the validation, and the software should make that boundary explicit.
Traceability per result
Every reported result has to be reconstructable: which pipeline version, which reference data, which parameters, which inputs. Not eventually, and not approximately.
Change control
An improvement to the analysis is a change to a regulated system. It carries assessment, revalidation where warranted, and documentation.
Locking the pipeline
Research pipelines improve continuously, and that is correct behaviour for research. A diagnostic pipeline is fixed at a version, with its tools, parameters, and reference data pinned, because the clinical validation applies to that configuration and not to a later one.
Changing anything inside it is a change control question before it is an engineering one.
The output is read by an oncologist making a treatment decision, not by a bioinformatician. What the result means, what was and was not assessed, and the limits of the assay all have to be legible on the report itself.
Ambiguity in a report is a clinical safety issue, not a presentation preference.
Our genomics work covers somatic variant calling and biomarker analysis using BWA, GATK, and SAMtools, and our digital pathology work quantifies protein biomarkers used in oncology treatment decisions. Companion diagnostics is where those capabilities meet the validation and traceability expectations of our regulatory practice.
Biomarker scoring models, an interactive slide viewer, Python and Playwright automation and end to end QA on a digital pathology platform for oncology.
Read the case study →