Are there really 345 risk genes for neuropsychiatric disorders?

ultifactorial diseases are the latest fad in biomedicine. A journal called Frontiers in Molecular Biosciences has a special issue on it.^[1] The editor writes

Multifactorial diseases arise from the dynamic interplay among genetic predisposition, environmental exposures, metabolic states, immune responses, and tissue-specific regulatory networks. . . . By combining genomics, transcriptomics, proteomics, metabolomics, and computational modeling, multi-omics approaches enable reconstruction of disease-driving molecular networks rather than isolated analysis of single factors.

Sounds great. There’s only one problem: there is no such thing as a multifactorial disease.

The term ‘multifactorial’ has two different meanings in science. The first is the idea that several individual events have to happen sequentially, as in cancer: the root cause is DNA and chromo­some damage, and only if all the many repair attempts fail does cancer occur. So what the editors above are talking about is not multifactorial diseases but cascading pathology.

This definition is fairly innocuous. Combining several techniques would be a good way of understanding what’s going on. It would also encourage labs in different institutions to collaborate more with each other. It’s just a misuse of the term: a chance to use a polysyllabic word that doesn’t actually fit.

The second meaning is the idea that a vast number of genes are ‘risk factors’ for a disorder. Thus, a disorder such as ASD (autism spectrum disorder) is said to result from any one of dozens, maybe hundreds, of defective gene products. The idea is that a disease might have no single cause but could result from one of several different things, each of which increases the risk slightly.

There are a handful of diseases that fit into this category. One example is FFI, or fatal familial insomnia. In FFI, a DNA mutation, or polymorphism, changes an aspartate (D) at position 178 of prion protein to an asparagine (N), so the mutation is called D178N. This mutation causes Creutzfeldt-Jakob disease (CJD), or mad cow disease. If the patient has a second polymorphism at codon 129, the patient no longer gets CJD but FFI. This causes them to lose slow-wave sleep and sleep spindles in their EEG. FFI patients either die in misery while fully conscious or they fall into a vegetative state and die from systemic or respiratory infection.

But for the most part, calling a disorder multifactorial is very misleading. The vast majority of diseases have a single cause. In these cases, if there is more than one cause, there is more than one disease and the problem is our inability to discriminate them.

A case in point is a recent study in Mol Psychiatry (paywalled, so no link). The authors ran a GWAS (genome-wide association study) and claimed that there are 345 genetic risk factors in neuro­psychi­atric disorders. The disorders were schizo­phrenia, Parkinsons, bipolar depression, MDD (major depressive disorder), ADHD, and ASD. The authors conclude that these disorders are all “multifactorial.”

It’s a typical example of how an ‘omic’ study can give us a lot of data but no information. The nicest thing you could say about findings like this is that they tell you to look elsewhere. The cause isn’t genetic, which is good to know. What’s really happening is that somebody has a big single-cell DNA sequencer. They have to justify the expense, so they do a study on it and then move on to something else. Journals will still publish papers using the method because it’s still fashionable.

In a GWAS, you take samples from hundreds of patients and sequence the entire genome. Then you calculate which genes correlate with the disorder. It is a well known problem that doing thousands of correlations on the same population always gives some genes that correlate by chance alone. Thus, a GWAS can never be definitive. Somebody has to follow up and confirm each of the hundreds of risk factors.

NINDS once published a program announ­ce­ment asking people to do just that for some disease to see if any of them are meaningful. It was a lost cause. I doubt that they got many submissions, or at least many fundable ones. If the people who did the GWAS believed their own results, they would have followed up on them themselves.

This is a little known fact that laymen don’t always understand: scientists often publish findings they don’t believe.

What the Mol Psych study really means is that we have no idea what causes the disorders. It is a negative result. The only thing we know for sure is that the GWAS did not tell us anything useful. The problem is that no author ever admits that. They always say the result shows the disease is “multifactorial”. That conclusion does not follow.

A GWAS always gives you something, and the statistical methods may be valid, but somehow every researcher gets a different, non-overlapping set of genes. I once compiled a list of the genes that various GWAS papers claimed for Alzheimer’s disease. Except for apoE, which we already knew about, there was no overlap—zero—among the genes. Many of them were expressed in the brain, yet they were all different. Every researcher ‘discovered’ a different set of genes, and every researcher claimed the experiment had worked. If only the authors would admit the truth—that they’d wasted a year of effort and half a million bucks on a bad idea—all would be well. But if they did that, they’d never get published. So they all claimed it meant that Alzheimer’s was “multifactorial.” That has now become the dogma.

Once a dogma spreads to the general public, it’s nearly impossible to dislodge. Big Pharma invents a drug that appears to work, people declare the problem solved, and research on the real causes slows to a crawl.

If something is multifactorial it means it is unimaginably complex. It’s really a way of saying it’s pointless to try to find the true cause.

If a disorder isn’t heritable, searching the DNA for polymorphisms is like looking under the streetlight for your lost keys. Take cancer: it is not a single disease but a vast collection of diseases that have something in common. If a GWAS was restricted to a single disease, it could be useful.^[2] But if not, as is happening in many disorders including ASD, schizophrenia, anorexia, and neuro­degen­erative diseases, the trend in science is to claim some improbably high number of ‘risk genes’, label the disorder multifactorial, and give up.

may 05 2026, 5:57 am