‹

11. The Uncharted Territory

EN

11. The Uncharted Territory

Two patients walk into two different clinics on two different continents. Both meet the same diagnostic checklist for major depressive disorder: low mood most of the day, loss of interest in activities once enjoyed, changes in sleep and appetite, difficulty concentrating, for at least two weeks. Both receive the identical diagnosis, written in the identical words on the chart. Brain imaging finds their underlying neurobiology does not match. This is not a hypothetical. It is a documented, published finding, and it is the starting fact of this chapter: one diagnostic label is currently being used to describe what may be several distinct biological conditions.

A 2020 paper by Dinga and colleagues in Neuropsychopharmacology used imaging-driven clustering — a statistical technique that groups patients by the similarity of their brain scans rather than by their symptoms — and identified multiple stable neuroanatomical subtypes hiding inside clinically identical depression samples. Later work, published in Biological Psychiatry between 2024 and 2025, extended the picture: two stable subtypes were found to differ significantly in cognitive ability despite reporting identical clinical symptoms on the standard checklist. The same diagnosis, arrived at the same way, can sit on top of measurably different brains. Nobody in this literature disputes that this heterogeneity exists. What remains unresolved — described in the source literature itself as "largely unclear" — is how to carve the messy, overlapping variation into stable, replicable categories that would let a clinician say, of any one person, which biological subtype they belong to.

This single fact does a great deal of explanatory work. It is, on its own, most of the reason why decades of biological psychiatry research have produced so many inconsistent, hard-to-replicate findings. A test that works well for one biological subtype of depression will look weak or noisy when run across a mixed sample that also contains several other subtypes it was never designed to catch. The label is one word. The underlying condition, or conditions, may not be.

No blood test

There is, as of the current literature through 2025, no validated blood test or biomarker panel that can diagnose major depression or reliably guide which treatment a given patient should try first. This is not an oversight or a funding gap. Multiple independent biomarker-discovery programs — looking at inflammatory markers, metabolic markers, neuroendocrine (hormone-system) markers, and broader multiomic panels that scan many biological signals simultaneously — have run through 2023 to 2025 without producing a marker or marker set that generalizes across the depressed population as a whole. Candidate signals do exist for subgroups of patients. But a signal that identifies a subgroup is not a diagnostic test for the disorder, and the heterogeneity described above is explicitly identified in the review literature as the reason no such test has reached clinical use: build a test tuned to one biological subtype, and it will simply miss the others hiding under the same diagnostic label.

Inflammation is the clearest illustration of this pattern, because it looked, for a while, like exactly the kind of unifying biological signature depression research had been missing. A 2019 meta-analysis by Osimo, Pillinger, Rodriguez, Khandaker, Pariante, and Howes, published in Psychological Medicine, pooled 30 studies measuring C-reactive protein, or CRP — a standard blood marker of systemic inflammation, produced by the liver in response to tissue damage or infection. Using the threshold of CRP above 3 milligrams per liter, the level the American Heart Association classifies as high cardiovascular risk, the meta-analysis found that 27 percent of depressed patients met that threshold for clinically significant low-grade inflammation. Using a looser threshold, CRP above 1 milligram per liter, 58 percent showed any elevation at all.

Read carelessly, that sounds like confirmation that depression is an inflammatory illness. Read correctly, it says something more limited and more useful: only about a quarter to just over half of depressed patients show elevated inflammation, depending on where the line is drawn — which means inflammation looks like a real marker for some depression, not a mechanism that explains the disorder as a whole. A 2025 commentary in the American Journal of Psychiatry frames this as an "inflammatory subtype" of depression rather than core pathophysiology shared by everyone with the diagnosis. Live debate in the field concerns whether inflammation, in the patients who show it, is a cause of their depression, a consequence of it, or simply a marker that happens to travel alongside a particular biological subtype without directly causing anything. The 27-and-58-percent numbers themselves are not in dispute. What they mean is.

The genes that were not there

The most decisive correction in this chapter concerns genetics, and it is worth telling in full because the study that produced it was designed, from the outset, to be unable to hide from its own answer.

For roughly two decades, depression genetics research concentrated on a short list of "candidate genes" — genes selected in advance, based on their known biological roles, as plausible causes of depression. Eighteen of the most historically studied candidates, including SLC6A4 (also known by its polymorphism name 5-HTTLPR, involved in serotonin transport), BDNF (involved in neural growth and plasticity), and COMT (involved in breaking down neurotransmitters), had each accumulated their own smaller studies, some reporting significant effects, especially in combination with environmental stress.

In 2019, Border, Johnson, Evans, Smolen, Berley, Sullivan, and Keller published a study in the American Journal of Psychiatry designed to settle the question with a sample large enough that a real effect, if one existed, could not hide. They combined data from the UK Biobank (up to 177,950 unrelated individuals) and the Psychiatric Genomics Consortium's genome-wide association data (443,264 individuals), for a total of 621,214 people. They tested all eighteen candidate genes, sixteen top polymorphisms, across eight different depression-outcome measures and five environmental moderators — including gene-by-environment interactions, the specific claim that a gene's effect on depression risk depends on exposure to stress. The whole design was preregistered on the Open Science Framework before the data were analyzed, meaning the researchers committed publicly to their methods and analysis plan before they could see which answer those methods would produce, closing off the option of quietly adjusting the approach after the fact to fit a preferred result.

Essentially none of it replicated. Only one gene, DRD2, showed a significant effect in one of the two datasets, and that single result failed to replicate in the other. The specific, famous claim that the gene 5-HTTLPR interacts with stressful life events to predict depression — the Caspi et al. finding that had anchored a large fraction of gene-by-environment research in psychiatry — did not replicate either, in a separate preregistered meta-analysis of 38,802 people cited within the same paper. Crucially, a power analysis showed the Border et al. study had greater than 99.99 percent statistical power to detect effects of the size the original candidate-gene studies had reported. A study with that much power finding nothing is not an ambiguous null result. It is strong evidence that the originally reported effects were never really there.

The paper's own stated conclusion was to abandon historical candidate-gene and candidate-gene-by-environment hypotheses for depression, a recommendation the authors note aligns with the National Institute of Mental Health's own Genomics Workgroup guidance. Earlier proponents of candidate-gene research, cited within the paper, had argued that prior replication failures reflected problems with how depression was measured or how environmental stress was defined, rather than evidence that the original findings were false positives — a defense this study's design, spanning eight outcome measures and five environmental moderators at massive scale, was specifically built to rule out.

The genome-wide association studies that superseded candidate-gene research paint a different, humbler picture of how depression risk is actually encoded. A separate 2018 study by Wray and colleagues, working with the Psychiatric Genomics Consortium's Major Depressive Disorder Working Group, examined 135,458 depression cases against 344,901 controls and found the single strongest common genetic variant, labeled rs12552, carried an odds ratio of just 1.044 — a barely detectable individual effect, requiring a sample of roughly 34,100 people to detect at standard statistical power (80 percent), even though the underlying statistical significance was extreme (p = 6.07×10⁻¹⁹, meaning the pattern was almost certainly not chance, even as the effect size itself stayed tiny). Depression's heritability, estimated from twin studies at roughly 37 percent, is real. It is simply distributed across many genetic variants of individually minuscule effect, not concentrated in a handful of "depression genes." This directly explains why the older candidate-gene studies, which had a median sample size of only 345 participants according to a review covering 2000 to 2009, were statistically incapable of detecting effects this small — and instead produced false positives at a striking rate: 96 percent of novel candidate-gene findings in that review reached statistical significance, compared with only 27 percent of attempts to replicate them.

No universal explanation, and no shame in that

None of this is a story of failure. It is a story of a field correcting itself in public, with preregistration, large samples, and honest reanalysis, which is exactly what functioning science looks like from the inside — messier than the tidy chemical-imbalance story ever was, and more accurate for being messier. What is missing is not evidence. It is a single mechanism, because there may not be one. One diagnostic label, several biologies, no gene that predicts it, no blood test that catches it, an inflammatory signature that applies to some but not most. The uncharted territory is not a gap waiting to be filled by one more study. It may be the honest shape of the thing itself.

Overturned if: a large, preregistered, adequately powered study identified a biomarker — genetic, inflammatory, or otherwise — that reliably diagnosed major depression or predicted treatment response across the general depressed population, rather than within one imaging- or biology-defined subtype.

Sources

บทถัดไป · 12 / 1212. Agency in a Predictive World
Size
Page
Font

The Architecture of Affect

❦
101. The Myth of the Broken Thermostat202. The Currency of Pursuit303. The Cost of Chronic Vigilance404. The Architecture of Resilience505. Rapid Restoration606. Relaxing the Predictive Constraint707. The Mathematics of Disappointment808. The Logic of Giving Up909. Environmental Recalibration1010. The Distributed Self1111. The Uncharted Territory1212. Agency in a Predictive World
‹ Prev11 / 12Next ›