Preprint Blood Immuno-metabolic Biomarker Signatures of Depression and Affective Symptoms in Young Adults, 2025, Donnelly et al

[forestglip]

Blood Immuno-metabolic Biomarker Signatures of Depression and Affective Symptoms in Young Adults

Nicholas Donnelly, Ruby Tsang, Èimear M Foley, Holly Fraser, Aimee Hanson, Golam M Khandaker

Background
Depression is associated with alterations in immuno-metabolic biomarkers, but it remains unclear whether these alterations are limited to specific markers, and whether there are subtypes of depression and depressive symptoms which are associated with specific patterns of immuno-metabolic dysfunction.

Methods
To investigate whether immuno-metabolic biomarkers could be used to profile subtypes of depression, we applied regression, clustering, and machine learning to a dataset comprising depression diagnosis, depressive and anxiety symptoms, and blood-based immunological and metabolic biomarkers (n=118). We measured inflammatory proteins, cell count, lipids, hormones, and metabolites from up to n=4161 participants (2363 female, 337 with depression) aged 24 years from the Avon Longitudinal Study of Parents and Children birth cohort.

Results
Depression at age 24 was associated with both altered concentrations of immuno-metabolic markers, and increased extreme-valued inflammatory markers. Inflammatory and metabolic biomarkers show distinct, opposing associations with somatic and anxiety symptoms. We identified two latent components representing the relationship between blood biomarkers, symptoms, and covariates, one characterised by higher somatic symptoms and inflammatory markers (neutrophils, WBC, IL-6), and the other characterised by higher anxiety and worry and lower inflammatory markers (CRP, WBC, IL-6). Individuals with higher somatic-inflammatory component scores had greater depressive symptoms severity over the next five years. Immuno-metabolic biomarkers predicted depression diagnosis (Balanced Accuracy=0.580) and depression with high somatic symptoms (Balanced Accuracy=0.575) better than chance, but not depression with high anxiety symptoms (Balanced Accuracy=0.479).

Conclusions
Alterations in immuno-metabolic homeostasis is present in young adults with depression well before the typical age of onset of cardiometabolic diseases. The relationships between affective symptoms and blood immunometabolic biomarkers indicate two biotypes of depressive symptoms (somatic-inflamed vs anxious-non-inflamed). These patterns are relevant for prognosis and prediction, highlighting the potential usefulness of immuno-metabolic biomarkers for depression subtyping.

Link | PDF (Preprint: MedRxiv) [Open Access]

 

[Yann04]

biotypes of depressive symptoms (somatic-inflamed

Makes me think about the fact that as far as I can tell in China ME/CFS is diagnosed as “Somatic Depression”.

 

[Utsikt]

Immuno-metabolic biomarkers predicted depression diagnosis (Balanced Accuracy=0.580) and depression with high somatic symptoms (Balanced Accuracy=0.575) better than chance, but not depression with high anxiety symptoms (Balanced Accuracy=0.479).

Balanced Accuracy is (Specificity + Sensitivity)/2

 

[Utsikt]

Figure 4: Performance of Machine Learning models predicting ICD-10 depression diagnosis and specific symptom profiles within individuals with ICD-10 depression.

The results are very unimpressive.

Edit: they should probably have mentioned the poor performance in the avstract. They only mention it briefly in the discussion:

Predictive performance was similar to models trained on sociodemographic data, and to brain-imaging models (Winter et al., 2024). However, performance was at an absolute level considered poor (Alba et al., 2017).

 

[rvallee]

I can't take seriously the idea of 'affective symptoms' having biomarkers of any kind, the whole idea is plain silly, and anyone who thinks of 'depression', if it is even one thing and not many, lumped together out of confusion, as mainly being 'affective symptoms', is being so extremely selective in their interpretation that there is no point even looking at the results. It's a level of selective reasoning and cherry-picking that would make even flat earthers go: "uh, this is a bit too much selection bias here".

The weird contradictions of this ideology are so absurd when you try to put them together in a coherent way. Always insisting that psychosocial processes are the main drivers, then imagining that it must, somehow, have biological impacts of some kind is just firmly in the same category of being unable to make sense as young earth creationists. There is the same appeal to magic and being driven entirely by conclusions.

What they're trying to do here, somewhat, makes as much sense as trying to study people's reaction to loved ones dying, and trying to separately study it as people's reaction to being told that their loved ones died, as if they are distinct things, as if the act of being told about such a tragedy can be separated from the tragedy itself. Let alone how anyone would go about studying this as it happens: "hey quick thing before, can you fill in those 15 questionnaires and hop in the machine while we take some of your blood real quick?"

This is one of the reasons why AI is going to be so revolutionary: the bias, the damn bias. Human bias is so damn extreme and everywhere all the time about all the things and taken so long past the point at which it's no longer reasonable. We'd have to create AIs that intentionally get stuck themselves in loops to get anything this foolish and inept. They'd have to want it. And sure enough they could be instructed to want those things, but they'd have to, they wouldn't be this bad on their own.