Scientists Finally Know What's Happening in Our Brains

For most of my adult life, the explanation I got for bipolar disorder was something like: your brain chemistry is off balance, these medications correct that, try this one and we’ll see how you feel in six weeks.

That was it. That was the science.

Not because doctors were lazy. It’s because that was genuinely all we had. Bipolar disorder was understood at the level of symptoms and observed drug responses, not at the level of what was actually breaking down biologically. We knew what happened. We had almost no idea why.

That is starting to change in a real way.

In the last two years, researchers have published what I’d call the most substantive cluster of bipolar science in history. A study with 2.9 million participants that mapped the genetic architecture of the condition in more detail than anything before it. A separate line of research that found specific proteins in brain fluid that can predict future manic episodes. And a new approach to brain stimulation that uses each person’s own brain map to guide treatment, rather than a generic protocol that works for the average patient.

None of this means the problem is solved. But it means we’re finally looking at the right things.

Here’s what the research actually says, written for people who live with this condition, not people who study it.

---

The biggest genetics study ever run on bipolar disorder

In January 2025, a consortium of researchers called the Psychiatric Genomics Consortium published a study in Nature that analyzed DNA from 2.9 million people, including more than 158,000 with a bipolar diagnosis. The goal was to find specific regions of the genome statistically associated with bipolar disorder.

They found 298 of them.

To understand why that number matters, you need to know that the previous largest study found about 64. This one found nearly five times as many. That’s not a small improvement. At that scale, you can start to see patterns across biological systems rather than just collecting data points.

What the researchers did next was more interesting than the number itself. They used those 298 regions to ask: which types of cells are most implicated?

The answer was neurons in the prefrontal cortex and hippocampus, specifically a class of cells called GABAergic interneurons. These are inhibitory cells, the ones that quiet down activity in a neural circuit when it needs to calm. If inhibitory signaling is what regulates mood states, this makes a lot of sense. Not in a vague hand-wavy way, but in a specific, mechanistic way that might eventually point to new drug targets.

There was one other finding I want to flag, because it’s genuinely strange and might matter more than it seems.

The same analysis found significant enrichment of BD-associated genetic variants in cells from the intestine and pancreas. Not the brain. The gut.

This is not the first time gut-brain connections have come up in mental health research, but this is stronger evidence than most prior findings. It raises the possibility that metabolic and digestive signaling pathways are not just incidentally affected by bipolar disorder but may be causally involved in mood episodes. That’s an entirely different frame for thinking about the biology.

---

Type 1 and Type 2 are not the same condition

This same study, and follow-up fine-mapping work from Hospital Clinic Barcelona, confirmed something that many clinicians have suspected but couldn’t prove: bipolar type 1 and type 2 have distinct genetic architectures.

Type 1 aligns genetically much more closely with schizophrenia, especially in dopamine signaling pathways. Type 2 aligns more closely with major depressive disorder and ADHD.

What this means practically: they’re not the same condition that looks different at the surface. They’re different conditions that share a name and some symptoms. The distinction matters for treatment because a patient whose genetic profile resembles type 2 may respond differently to antidepressants than someone with a type 1 profile. And vice versa with antipsychotics.

Fine-mapping work narrowed the 298 regions down to 36 specific genes with high confidence as causal. Several of them encode parts of voltage-gated calcium channels, which is exactly the mechanism that existing mood stabilizers like lithium are believed to affect. So the genetics are converging on something that the pharmacology already pointed toward, which is a good sign.

---

What’s actually in the brain fluid

Genetics tells you about predisposition. Proteomics tells you what’s actually happening right now.

A 2025 study published in Biological Psychiatry did something that hadn’t been done at this scale before: they took cerebrospinal fluid (the fluid that surrounds the brain and spinal cord) from 374 people and mapped more than 2,000 proteins in it. Then they compared what was present in people with bipolar versus those without, across two independent groups, to make sure the findings were real and not flukes.

They found 41 proteins that showed up consistently different. Here’s what the pattern looked like.

Lower in bipolar disorder:

• NRXN1 (Neurexin-1), a synaptic protein essential for maintaining connections between neurons
• NPTX2 (Neuronal Pentraxin-2), involved in synapse formation
• CLSTN1 (Calsyntenin-1), linked to memory and learning circuits
• Several axon guidance molecules that help neurons maintain their architecture

Higher in bipolar disorder:

• Complement proteins C3 and C5, markers of immune activation in the brain
• Blood-brain barrier integrity markers

The picture that emerges is a brain with reduced synaptic connectivity, fewer of the proteins that maintain strong neural connections, and elevated inflammation markers that suggest the immune system is involved in what’s happening. This is biologically coherent with what’s observed clinically. It doesn’t mean bipolar is an inflammatory condition in the simple sense, but it suggests the immune system isn’t just a bystander.

The part that changes everything for tracking

The study followed participants for a median of 6.5 years. Here’s the finding that should matter most to anyone building or using a tracking system:

Changes in those protein levels over time correlated with whether people had manic episodes later. Not just whether they had worse symptoms generally. Whether they had future manic episodes, specifically.

CSF testing is not something you do at home. But the behavioral signals that map to those same underlying biological systems, sleep variability, social rhythm disruption, activity changes, are trackable today. The fact that your sleep patterns can signal an episode 1 to 6 days out is not a coincidence. It’s the surface expression of biology that researchers can now describe at the protein level.

Your data knows before you do because the underlying biology moves before your conscious experience catches up. Now we know more about what that biology actually is.

---

Bipolar isn’t a broken region. It’s a network problem.

For a long time, neuroimaging research tried to find the one region of the brain that was abnormal in bipolar disorder. Some studies pointed to the prefrontal cortex. Some pointed to the amygdala. Some pointed to the hippocampus. None of it was consistent enough to be clinically useful.

The more recent framing is different, and I think it’s closer to the truth.

Bipolar disorder is a problem of large-scale network dysfunction. Not one region. Three interconnected networks that talk to each other constantly and stop doing so normally.

The Default Mode Network handles self-referential thought, what you ruminate on when nothing demands your attention. The Salience Network decides what’s worth paying attention to and what counts as a threat. The Central Executive Network handles working memory and cognitive control, the ability to hold competing priorities in mind and make decisions.

In bipolar, these networks show reduced connectivity. The right conversations between brain regions aren’t happening. The Salience Network miscalibrates emotional significance. The Central Executive can’t hold things steady.

This is a meaningful shift in understanding because it explains why the illness feels the way it does. The instability isn’t localized. It’s systemic.

---

New treatments that actually follow from the science

Personalized TMS

Transcranial Magnetic Stimulation uses magnetic fields to stimulate specific regions of the prefrontal cortex. Standard TMS targets a fixed scalp location based on population averages. It works for some people and not others, and no one has been great at predicting who.

Researchers at Weill Cornell and Stanford are doing something different. They take each patient’s own resting-state fMRI, map which subregion of their prefrontal cortex is most functionally connected to the subgenual cingulate, a key node in mood regulation, and then target that specific location for stimulation. The target is personalized to the individual’s actual brain map, not an average.

Early trials show better outcomes than standard protocols. This won’t be available everywhere for years, but the direction of travel is clear: treatment based on your biology, not a population average.

Accelerated TMS

Standard TMS requires daily sessions for 4 to 6 weeks. That’s a serious commitment when you’re depressed. Getting out of bed is already hard; getting to a clinic every day for a month is harder.

Accelerated protocols run multiple sessions per day and compress the whole course into 5 days. Clinical trials in 2025 show outcomes comparable to standard protocols. If those results hold, this becomes a meaningfully more accessible option for people in acute bipolar depression.

Ketamine for bipolar depression

Ketamine works on the glutamate system rather than the dopamine or serotonin systems that most existing mood medications target. It acts in hours, not weeks. For someone stuck in a bipolar depressive episode, that speed matters.

Multiple clinical trials are evaluating ketamine specifically for bipolar type 2 depression as distinct from unipolar depression, which is the right approach. They’re not the same condition and shouldn’t be treated identically. Results so far are promising enough that several centers are running formal protocols.

---

What this means for tracking your own data

The research above points to one conclusion for anyone managing bipolar day to day: the signals worth tracking are the ones that map to underlying biology, not just the ones that feel significant.

Sleep and circadian rhythm disruption show up directly in the 2025 genetics findings. Several of the 298 identified loci involve CLOCK and ARNTL, core genes of the circadian system. Bipolar disorder is, in part, a circadian rhythm disorder, and the genetics confirm it.

Social rhythm and activity levels are behavioral proxies for the synaptic protein disruptions found in the CSF study. When NRXN1 is low, synaptic connectivity degrades. When synaptic connectivity degrades, you behave differently, you withdraw, your patterns shift, your social interactions change.

Tracking irritability is tracking the early surface expression of the salience network misfiring. When your brain starts miscalibrating what counts as a threat, small things feel disproportionately significant. That irritability shows up before everything else, because the network disruption happens before it propagates to full mood changes.

The point isn’t that your app is a medical device. It’s that the behavioral signals you can measure daily are genuinely connected to the biology researchers are now mapping. Tracking them isn’t pseudoscience. It’s working with the grain of what’s actually happening.

---

Where this leaves us

Nobody is cured. The 298 genetic loci don’t translate into a pill. The CSF proteins can’t be measured at home. The personalized TMS isn’t available at most clinics. The gap between research and clinical reality is still large.

But the framing has changed in a way that matters.

For most of the history of psychiatric treatment, bipolar disorder was managed through symptom observation and drug trial and error. The biology was a black box. Now there are specific proteins, specific genes, specific network dysfunctions, specific cell types that are being mapped with real precision.

That means better treatments are coming, eventually, because researchers are finally solving the right problem. And it means the instability you’re managing is not random or character-based or vague. It has a specific biological shape.

That’s worth knowing.

---

Sources: Nature, January 2025 (PGC GWAS, 298 loci); Biological Psychiatry, 2025 (CSF proteome, Goteson et al.); Hospital Clinic Barcelona, 2025 (36 causal genes, fine-mapping); USC/ENIGMA-BD, 2025 (brain imaging consortium); BD2 Initiative, 2025 (personalized TMS)