Interpretable classification for point-of-care diagnostics: chemometrics partner

Datharsis, a University of Granada spin-off specialised in explainable data analysis, is looking to join a consortium as the partner responsible for signal and data analysis, either leading a work package or contributing specific tasks.

A point-of-care device is only as good as the model that turns its raw signal into a result. In resource-limited settings, with small reference sets, noisy signals and variation between devices, operators and batches, the model must be robust, validated and explainable, because a diagnostic decision has to be defended before clinicians and notified bodies. Chemometrics on spectral and sensor data is where our methods come from.

What we offer:

• Design of reference sample sets: which samples, from where and how many, to build a reliable model

• Interpretable classification and calibration models on spectral and biosensor data (FTIR, NMR, Raman, electrochemical signals), showing the variables behind each decision

• Handling of instrument, batch and operator variability, separating it from the diagnostic signal

• Rigorous validation with cross-validation and permutation testing, to quantify real diagnostic performance

• Reproducible, auditable pipelines that support regulatory documentation

• Access to DathaSuite, our analysis platform, and training for consortium staff

Our team’s methods have been applied to classification and authentication problems on spectral data, and to biomarker discovery in clinical studies. See the documented cases.

We are looking for consortia with device developers, biosensor groups, analytical chemistry labs or clinical validation sites that need the data analysis side covered.