Gaps, cross-references & frontiers

A critical map of what we don't know, the cross-domain connections no one has tested, and the research running right now. The opportunity: depression science is rich in single-exposure associations and poor in causal, interaction-aware evidence.

The biggest gaps & contradictions

Field-wide problems: cross-sectional dominance, pervasive reverse causation, publication bias, self-report error, functional unblinding, small samples, weak animal translation, diagnostic lumping, and WEIRD-sample bias (the evidence base is overwhelmingly Western while burdens are highest elsewhere).

Unexplored cross-references (testable now)

Microplastics → gut → kynurenine

Dietary microplastics may shift the microbiome to divert tryptophan to neurotoxic kynurenine. Test: fecal microplastics + 16S + plasma kynurenine/tryptophan + PHQ-9.

PM2.5 × genetic risk → kynurenine

A confirmed gene×pollution interaction (OR≈3.2) may be mediated by kynurenine. Test: UK Biobank PM2.5 × PRS × metabolomics.

Ultra-processed food as a bundled vector

Emulsifiers + microplastics + low fibre + glycaemic load + aluminium may jointly raise IL-6. Test: composite UPF index + IL-6 + PHQ-9.

Cadmium as the hidden smoking mediator

Much of smoking's effect may run through cadmium, not nicotine. Test: NHANES cadmium + selenium + PHQ-9 mediation.

PFAS × thyroid × circadian

PFAS suppress thyroid hormones that entrain clock genes. Test: NHANES PFAS + free T4 + actigraphy (all variables already exist).

Fatty liver (MASLD) → brain

UPF-driven liver inflammation may cause an anhedonic "metabolic depression." Test: UK Biobank liver-fat MRI + PNPLA3 Mendelian randomization.

What's running now — the pipeline

Do first: the four "desk-ready" interaction analyses above (NHANES + UK Biobank), adopt CRP-high/low stratification everywhere, and pre-register with held-out replication before any AI interaction-mining. Full detail: download the Frontiers report → · Deep Dive: verdicts, runnable protocols & the convergence verdict →