Blood markers, muscle loss, and AI dependency walk into a clinic.

Two blood proteins may help flag bipolar disorder — with a catch.

What if a blood test could one day tell a psychiatrist that your brain is under oxidative siege — before you even describe your symptoms?

A team published in BMC Psychiatry measured a panel of molecules in the blood of 101 people with bipolar disorder and 29 healthy controls. Two of those molecules — TrxR1 and PRDX1, both proteins your cells use to mop up internal damage — were significantly elevated in every bipolar patient, regardless of whether they were in a manic episode, a depressive one, or a calm period in between. A third molecule, GSH (a natural antioxidant, think of it like a household cleaning spray your cells produce), was significantly depleted in patients. Here's the plain-English version of what's happening. Your cells constantly generate byproducts — call them exhaust fumes — from normal metabolism. Healthy cells neutralize these quickly. In people with bipolar disorder, this cleanup system seems to be chronically overloaded. TrxR1 and PRDX1 are essentially the cleanup crew working overtime. Finding them elevated in blood could, in theory, serve as a marker the way checking your car's oil level tells you the engine is under stress. The catch — and it's a real one — is that both proteins were elevated equally across all three mood states. So they don't tell you *which* episode someone is in. They can't replace clinical assessment. This is also a cross-sectional study, meaning everyone was measured once; we don't know if the levels fluctuate meaningfully over time. The sample is 101 patients from one hospital. And the paper doesn't fully detail how the team corrected for multiple statistical comparisons, which matters. What this is: an early signal that something biological and measurable is chronically dysregulated in bipolar disorder, full stop.

In schizophrenia, muscle loss and inflammation each make symptoms worse — separately.

Nearly 1 in 5 people with schizophrenia in this study had significant muscle loss — and it made their psychiatric symptoms measurably worse.

Researchers studied 120 people diagnosed with schizophrenia and found two overlapping but independent physical problems: 17.5 percent had sarcopenia — that's the gradual, pathological loss of muscle mass and strength, not just being thin — and 34.2 percent had elevated CRP, a blood marker that goes up when the body is inflamed. Think of these as two separate cracks in the same foundation. The interesting part is what each crack does on its own. Patients with elevated inflammation had higher scores on the PANSS — the standard scale psychiatrists use to rate how severe schizophrenia symptoms are — and they also performed worse on a test called the WCST, which measures cognitive flexibility (your brain's ability to stop doing something when the rules change). Patients with sarcopenia also had higher PANSS scores, but their cognition scores didn't differ significantly from those without muscle loss. So inflammation appears to hurt both symptoms and thinking. Muscle loss appears to hurt symptoms but not cognition, at least not in the ways this team measured. These are distinct pathways, not the same problem wearing two masks. The catch: this is cross-sectional, meaning we took a snapshot, not a video. We cannot say whether sarcopenia caused worse symptoms, or whether being more severely ill makes you less likely to move, eat well, and maintain muscle. The causation is genuinely unclear. What we can say is that 1 in 5 patients in this sample had a measurable physical condition that most psychiatric assessments don't even look for. That alone is worth paying attention to.

Students who doubt themselves academically stress out more — and lean harder on AI.

The less confident you feel in your own schoolwork, the more stressed you get — and that stress is what drives you to offload tasks to AI.

Researchers at Universitas Negeri Surabaya surveyed 274 students from a single program and asked them three sets of questions: how confident are you in your academic abilities, how stressed do you feel about school, and how much do you rely on AI to do your work? They then ran a statistical model — PLS-SEM, which tests chain-link relationships between multiple variables at once — to see what predicts what. The chain they found goes like this. Low academic self-efficacy (your belief in your own ability to handle coursework — not raw intelligence, just confidence) leads to higher academic stress. That stress, in turn, leads to more AI dependency — meaning using AI not as a tool you choose, but as a crutch you reach for because you feel you can't manage without it. It's like a student who, unsure they can build a shelf themselves, eventually just orders it pre-assembled every single time. The doubt doesn't solve itself; it becomes a habit of outsourcing. This matters for mental health because academic stress is a well-established predictor of anxiety and burnout. If AI dependency is partly a symptom of that stress rather than a neutral tool choice, then interventions aimed purely at 'responsible AI use' may be missing the point. The honest limits: this is one program, one university, one country, one moment in time. The statistics confirm a pattern but cannot confirm cause and effect. It's also possible that using AI for hard tasks is perfectly rational, not pathological — the paper doesn't fully settle that distinction. Treat this as a provocative early signal, not a verdict.

Three papers, three different angles — and yet they're all pointing at the same uncomfortable idea: the body and the environment are doing things to mental health that pure psychology keeps underestimating. The bipolar biomarker study says the brain's crisis is also a cellular one, detectable in blood. The schizophrenia study says psychiatric wards may be routinely missing a physical condition — muscle loss — in roughly 1 in 5 of their patients. And the AI dependency study says academic stress isn't just a feeling; it reshapes behaviour in ways that might entrench the problem. None of these papers is conclusive on its own. But if you squint at them together, the picture that emerges is this: mental health treatment that focuses only on mood, thought, and talk may be leaving significant leverage on the table — in the body, in the classroom, and in the tools we hand students without asking why they're reaching for them.

The bipolar biomarker story is only interesting if future work can show these markers fluctuate across episodes in the same patient over time — a longitudinal study would transform this from 'interesting signal' to 'potential diagnostic tool.' On the schizophrenia-sarcopenia front, watch for trials that add physical rehabilitation to psychiatric treatment and measure symptom outcomes. That would be the real test. And as AI tools become more embedded in education, we should expect — and demand — more studies that ask not just 'are students using AI?' but 'what are they feeling when they reach for it?'

• Wikipedia: Oxidative stress — what it is and why cells care
• Wikipedia: Sarcopenia — the gradual loss of muscle mass
• Wikipedia: Self-efficacy — Bandura's concept explained