Biomarkers Are Not Drug Targets, and Why That Distinction Costs the Industry Billions

If one assumption has quietly cost pharmaceutical R&D more than any other in the past two decades, it is the conflation of biomarker with drug target. They look similar on a slide. They diverge sharply in clinical trials.

A biomarker is an indicator — a measurable signal correlated with disease presence, activity, or progression. A drug target is different: a molecule, pathway, or cell state that is causally involved in the disease and can be intervened upon to change its course. The two are not interchangeable, and treating them as if they are has been measured in failed trials, abandoned programmes, and lost decades.

The TNF-Alpha Cautionary Tale

Take TNF-alpha. It is a beautifully reproducible biomarker in inflammatory disease and the target of a generation of blockbuster inhibitors that — despite commercial success — leave many patients only partially treated, expose others to serious adverse effects, and fail entirely in certain inflammatory conditions where TNF-alpha is elevated. The biomarker is real. The causal pathway is more complicated.

Why This Happens

First, mRNA fold changes do not reliably translate to protein changes: bulk transcriptomics can flag a gene as upregulated tenfold while protein abundance barely moves. Second, biomarkers are relative, not absolute — overexpression in one disease does not mean absence in related conditions, so trials enriched purely on biomarker positivity enrol biologically mixed populations and wash out treatment effects. Third, bulk omics average across cell types: a signal that looks important at the tissue level may originate from a single subpopulation and be irrelevant elsewhere.

The Way Forward

The way out is conceptually simple and methodologically demanding: stop treating biomarkers as targets, and start identifying the causal cell states that drive disease. Single-cell omics has, for the first time, made this practical at scale. We can ask which specific cell populations, in which transient states, are doing the pathological work; whether those states are druggable; and how to intervene at the right cellular level and the right disease stage.

This is not a niche academic point. It is the structural reason so many "rationally designed" trials fail despite good biomarkers, good chemistry, and good intent. The next generation of successful programmes — in oncology, immunology, and critical care — will be built on cell-state pharmacology, not biomarker pharmacology. If your programme rests on a biomarker-led hypothesis, it is worth asking: which cell, in which state, at which moment in disease, is the drug actually supposed to change? If you can answer that, you have a target. If you cannot, you have a publication risk and a phase-3 problem waiting to happen.