Pinning down disease susceptability genes is not working? Say it not so!

The headline reads:

The Human Genome Project failed to deliver

I always hope that people aren't surprised by this kind of thing but that hope is always to be dashed it seems.

This elicits the same bit of "duh, what part of self-evident weren't you getting" that came from mind-numbingly stupid "1-Gene-1-Trait" (or equivalently "1-Genotype-1-Phenotype") model almost ten years ago.

I was working in a genomics company back when that model was smashed by the results of the human genome project. I was shocked that any PhDs in biology were even surprised. Some had an intuition it was wrong without a rigorous basis. Good for them! Most were stunned. It was sad. But apparently "the false meme still lives on".

There are only rarely going to be "a gene for X" in genomics. It will be so rare that it's really foolish to start from that presumption.

The mind-bogglingly simple reason being that there is a complex (nonlinear) cyclic graph network between genotype (DNA and RNA expression) and phenotypical expression called the metabolome (the metabolic pathways).

This very much akin to electrical circuits connecting a radio antenna to a loudspeaker in your radio. You don't expect to find a single "Britney Spears" radio station as a singular cause of hearing "Britney Spears" songs on all over the dial.

First stations don't play only Britney Spears - presuming that is missing a deep understanding of why and how a station functions as it does. By analogy, diseases aren't generally caused by the failure of a gene, even most of the time. In fact, it's pretty darn rare that *any disease* will be directly genetic that way. The 1:1 mapping is wrong - radio station != music heard. They are related but are not the same thing and they are not mathematically bijective or injective or directly surjective. The topology is more complex than that.

Second, you can't presume some simple-minded model of a "Britney" station somehow "being active" and forcing itself through the circuit network causing Britney to appear on the speakers when you spin the dial. Yet presuming an "X gene" is pretty much the same thought process. It's moronic!!

When it comes to genes, most diseases are failures of the network between genotype and phenotype! This is why metabolomics is so important and also why it will take much longer to understand, and even why it's possible some things will never be predictable. Genes can trigger such failures but genes are not even necessary causes. I am quite sure that we'll discover many of the so-called "disease genes" already claimed will in fact turn out to be structural genes not in any way related to the disease. Something like blaming the concrete or steel "gene" of a building for allowing the disease of collapse actually caused by an earthquake or hurricane. The current attribution will be seen as a laughable case of completely missing the point.

This network is cyclic, nonlinear and interconnected. Having cycles, with or without nonlinearity, creates more than enough complexity to account for the "1-genotype-1-phenotype" discrepancy and probably creates enough on top as well to keep biologist employed for the next millennium. It's call feedback, especially nonlinear feedback. With some basic math it's easy to see why there's probably plenty of complexity to account for "too many traits and too few genes".

The good news is some parts of the metabolome have been found to be either scale-free or nearly scale-free rather than perfectly random like an Erdos-Renyi network. That means the 80-20 rule should apply to a certain extent: 80% of activity will be confined to 20% of the network.

However that last 20% has been conserved for a reason, and likely we'll need to get round to that part to understand 80% of disease which is, by definition, exceptional (like 20% is exceptional). Likely that effort will face by the same 80-20 "rule" which assures that it takes 80% of the time to uncover.

The really nasty bit is that until all biologists start learning calculus and differential equations as a standard part of an undergraduate degree (akin to engineering or physics), biologists will never understand nor create a workable model for the interplay of genomics, expression, regulation, metabolomics and disease. Never. No way. No how. That's simply the nature of what comes with such networks. The good news is they can borrow heavily from what engineering has already embraced from physics and mathematics.

There's a very good chance biology will either be forced to abdicate this realm or split into "soft" and "hard" biology akin to what bioinformatics has already been doing. I'm hopeful they will simply adopt more math rigor across the board because it isn't just molecular biology that would benefit, and that is currently held back by lack of math. Ecological and systems levels biology are suffering and stagnating in similar ways.