Isn't your DNA largely insufficient for determining phenotype, particularly locally as in the case of cancerous cells? It takes a tremendous amount of chemical and software processing to capture a snapshot of quantitative protein expression, and even more to measure, say, phospho-regulated protein expression, but that is what is really needed for diagnosis and tailored treatment. You may be right about the "precipice", but I feel like you are over-selling pharmacogenomics.
Definitely for certain things like intelligence, but many disease processes will be one gene coding one deleterious enzyme. For example Hyper-IgE syndrome will be a STAT3 mutation, Cystic Fibrosis is going to be a change typically at 508 to F(amino acid phenylalanine) on the CFTR gene, and Menkes disease, with its kinky hair, is going to have a mutation with ATP7A. These phenotypes are easy to recognize because the symptoms precipitate so significantly that most individuals die prematurely and because we've isolated them time and time again in the same diseased individuals. My point here is that isolating a phenotype from a genotype is pretty easy for a significant amount of polymorphisms.
Now what makes drug metabolism easy to understand is that it will typically be one gene that codes for one enzyme (I'm overly simplifying this here, it's way more complicated and we don't fully understand gene expression). It will also usually be one enzyme that is going to break down a certain drug (I'm also oversimplifying this). Let's look at ethanol; it's broken down by alcohol dehydrogenase to acetaldehyde, the thing that gives you the hang over. Certain populations, Asian and Native Americans namely, have a different alcohol dehydrogenase enzyme than other populations and thus have a common effect; colloquially termed Asian flush. This variability in how an enzyme breaks down a drug is part of the reason why some people will have certain effects whereas other people won't (overly simplified, again). Pharmacogenomics is actually one of the more easy phenotypes to delineate versus other, more nebulous, things.
Hmm. I think you are over-selling genomics as a tool for significant future advances in personalized medicine because it is only useful in a limited number of straightforward cases (as you say: easy phenotypes). After all, an individual's DNA contains only most basic information. I think a more likely source for advances in personalized medicine is from live snapshots of an individual's proteome. Genomics is like reading the source code and proteomics is more like a debugger.
DNA not only contains the most basic information, but all the information. It houses more information than what we get purely from protein expression. Where your point falls short is that we don't need to know every nuance behind protein expression, ubiquitination, posttranslational modifications, etc. to know that if patient A has polymorphism X and polymorphism X means that they cannot have succinylcholine, because they are susceptible to malignant hyperthermia or hyperkalemia we will not give them succinylcholine. My next point is that you will have variable expression of proteins based on drug concentration (the lay meaning: whether we have administered a drug or not). The classic example of this is the Lac operon, which is taught even in medical school. It demonstrates that when lactose is present the enzyme lactase will be made to metabolize it, but only when lactose is present. If a drug is not present its enzyme could not be as well skewing results in a way that could lead the physician to not try first line medications. Why would I want a snapshot of something when I need to predict drug-gene interaction before I give someone a drug?
Yes, pharmacogenomics is a home run on straight forward cases, but the amazing thing is that straight forward cases are the most important for medicine. Drug-gene interactions are so unbelievably basic and require next to zero know how to correlate side effects to genes. My previous example of genomics not being lucid is in intelligence because we have zero idea what makes someone intelligent on a molecular level, which physicians could not care less about.
My final point is purely on practicality. Last I had heard, there are some hospital systems that have genome sequencing are down to tens of dollars and are done in under an hour or two. Isolating a single protein, looking at its expression, and assaying activity is not only the bane of every biochemists existence, it takes a significant amount of time and would have to be done for every single enzyme that we want to look at. We have to be practical in medicine and the most practical option for the foreseeable future, unless a CRISPR like breakthrough comes along, is going to be genome sequencing.
Now, what is going to be absolutely insane in the field of proteomics is when we can accurately predict a protein's structure and then make novel drugs to enhance or inhibit said protein. I can't remember how many amino acids it is before we're unable to predict the structure of a protein, but it's low and the current method to look at protein structure is stupid hard. Obnoxiously hard; I don't even want to talk about it but it's hard and you have to send your samples to a handful of locations. We use super computers at the moment to elucidate small proteins and usually they suck bad at doing it. If we were able to accurately predict structure and function of a novel protein with computer models, use CRISPR to insert the gene into a genome for said protein we are talking about increasing human life span in a logarithmic fashion.
> DNA not only contains the most basic information, but all the information.
This is simply false. I don't understand why you would make this claim. You mention counterexamples below (e.g. PTMs). As you know, DNA dictates whether a protein can be expressed or not, but not whether it is expressed in any particular cell, how much is expressed, or whether and how much is activated or otherwise modified over time. And, of course, it can't contain environmental factors, such as the gut microbiome from the OP.
> Isolating a single protein, looking at its expression, and assaying activity is not only the bane of every biochemists existence, it takes a significant amount of time and would have to be done for every single enzyme that we want to look at.
I worked for a few years in a lab doing quantitative protein expression, including certain PTMs, that can compare multiple states (e.g with and without a drug, or in different tissues) across a large section of the proteome. This was not particularly new 10 years ago, but it is a lot of work, and sensitivity, specificity, and practicality is still improving rapidly.
> Why would I want a snapshot of something when I need to predict drug-gene interaction before I give someone a drug?
Because there are cases (probably the majority) where drug-gene interaction is insufficient.
I'm certainly not trying to argue that there aren't significant examples where the gene existence is sufficient. It sounds like that will continue to be an essential part of diagnosis and treatment. It is not surprising that current state of the art stops at these straightforward gene-based analyses, but it is surprising that you think it will be sufficient going forward and don't think the nuance of real protein expression will be useful. There is so much room between measuring "gene-existence" and measuring "intelligence". Adding pharmacoproteomics to your list of "ways 21st century medicine will be awesome" is a pretty gentle request!
If DNA doesn't code for everything then you deserve the Nobel prize because there isn't any other mechanism for how cells derive structure and function from that is known outside of this conversation. What else codes for PTM, ubiquitination, and basal level expression.
>DNA dictates whether a protein can be expressed or not, but not whether it is expressed in any particular cell, how much is expressed, or whether and how much is activated or otherwise modified over time.
It does determine whether it is expressed and how much is expressed. The difference between being a heterzygote and homozygote for certain proteins literally determines whether one will have fully protein expression or half. When a gene is doubled in the genome its expression is usually doubled as well.
>Because there are cases (probably the majority) where drug-gene interaction is insufficient.
80-90% of drugs are metabolized by CYP2D6, CYP2C19, CYP2C9, CYP3A4 and CYP3A5 to which we are starting to isolate polymorphisms in the general public. This isn't that difficult to do the genotyping, but what's limited right now is correlating polymorphism to what the side effect of drug is; this is going to be purely a time and data collection issue. I'm curious what your examples would be, because the majority of cases I've been apart of it's a simple base pair substitution that cause deleterious effects.
What is some open source software that I can contribute to in order to make these easier? I'd prefer your recommendation for something actually used/promising, instead of simply searching github. Thanks
