Single-analyte tests collapse heterogeneous biology into one number. High-dimensional panels recover the structure that stratification depends on — if the analysis is designed before the data is collected.
A diagnostic is a compression of biology into a decision. The question is how much structure survives that compression. Single-analyte neurology markers are reproducible and cheap, but they average across mechanisms that respond very differently to intervention — which is one reason trial cohorts assembled on symptom labels so often dilute a real effect.
What high-dimensional panels recover
Multi-parameter cytometry and targeted transcriptional panels describe the neuronal, glial and neuroimmune compartments simultaneously. The value is not the additional markers as such; it is the covariance between them. Two samples with identical inflammatory marker levels can occupy entirely different regions of a high-dimensional space, and those regions predict different responses.
- Panel design fixes the questions that can be asked later — design it against a hypothesis.
- Batch structure must be randomised across biological groups or it becomes the dominant signal.
- Donor-matched controls on every run make cross-cohort comparison possible.
From clusters to stratification
Unsupervised clustering produces groups; stratification requires that those groups be stable, interpretable and reproducible in an independent cohort. We treat any cluster that does not survive a held-out cohort as an artefact of the analysis pipeline, not a patient subtype.
Analytical validation before clinical claims
Precision, linearity, limits of detection and operator variance are established before any biological interpretation is published. A biomarker that has not been characterised analytically cannot be said to have failed or succeeded clinically.
This article describes preclinical research. Cytosomatic Labs does not treat patients, sell products or provide medical advice.