Look forward to the rest of this series Elliot! Especially curious to hear your thoughts on DNA synthesis, and how enzymatic approaches might evolve over time.
I might also add that beyond breakthrough technologies enabling us to understand how our genomes impact health, our ability to make meaning out of this new information is critically important. I shared some brief thoughts on this today: https://healthandwealth.substack.com/p/genetic-counseling
Great post! Given the general interest focus of this series, I hope you won't mind some noob clarification questions (coming from someone whose last exposure to Biology was AP Bio in the early 00's) about the
"[Sequencing data] will have uncontroversial benefits like finding better drug targets for diseases"
- Can you provide an example?
- If we've already sequenced a bunch of people, can you give some intuition around how and why sequencing many more will lead to finding better drugs for certain diseases?
"[Sequencing data] will greatly improve precision medicine-where patients are matched with drugs based on their genetics"
- Can you give an example?
- How can we learn this relationship? Even when the health outcomes are easy to measure (not the case in mental health, for instance) how exactly would we be able to detect that certain drugs are more effective for people certain genes? From a statistical standpoint, if there are dozens of genetic subpopulations that may respond differently to dozens drugs for a given disease, doesn't that lead to an explosion of treatment groups during clinical trials?
"Sequencing will also be used more directly in clinical care," and the two examples you give are diagnostic.
- Can you explain the first example? (What's a causal mutation?)
- In the simplest case, I understand how cheap sequencing could detect things like congenital single gene diseases. And though I don't understand what polygenic risk scores measure, exactly, I understand being able to say "people with a genetic makeup similar to yours go on to develop [insert disease here] x% of the time." And how that would lead you to take preventive steps to addresses [insert disease]. So, given that we already do seem to do this when people's lifestyle or demographics are similarly predictive of [insert disease], how much extra predictive power does sequencing give us?
- Going beyond "you were born with these genes, and we know that causes or is at least a risk factor for [insert disease]," the cancer diagnosis example seems to imply "we can detect disease before it fully develops because it's changed your DNA." Are there any other diseases that we can detect in this "non-congenital" way?
"[Sequencing gut microbes] could become essential for understanding how different diets actually impact our bodies on an individual basis." The dismal state of nutrition science suggests that we already have a hard time measuring the relationship between diet and impact on our bodies (likely because there are so many input variables, and the interesting outcomes take years if not decades to manifest). Would adding the additional variable of gut micro composition somehow make this topic easier to study, instead of harder?
Awesome, thanks for reading! Great questions. Some answers:
- On average, drug targets that are supported by statistical patterns discovered in human genetics are more likely to succeed. Based on this, there has been a big shift in biopharma to use more genetics in drug discovery. The UK Biobank project I mentioned is actually a public-private partnership with several pharmas using the findings for drug discovery. More on this here: https://doi.org/10.1038/s41588-021-00885-0.
- Sample size is really important for human genetics, because some important rare variation only happens at low frequencies in the population. Because ancestry is important, we also need to do a much better job of sequencing non-European populations. One of the preprints I linked to does a nice job illustrating this: https://doi.org/10.1101/2022.01.07.475305 as you sample more you start to saturate the total predictiveness of genetic information, but it is on a population-specific level (here only for Europeans).
- A person's genetics can impact how they will respond to a drug. This is true in cancer, where only certain mutations are treatable with tyrosine kinase inhibitors. It's more broadly true than cancer though, for example your HLA haplotype determines how well you will respond to certain antiretroviral drugs for HIV. There's actually a whole field working on this type of problem called pharmacogenomics, with many types of interesting stories in gene-specific drug responses: https://en.wikipedia.org/wiki/Pharmacogenomics
- The answer to your second question is linked with the first. Making these discoveries requires large genotyped drug response databases, and is still an active area of research. It typically involves finding how certain genotypes differentially metabolize drugs.
- A causal mutation is a sequence change in a gene that is causing the observed phenotype—in this case an inherited disease. So the key goal in clinical genetics is to identify this mutation to solve the case. In this case, it was an undiagnosed patient with an unexplained heart defect. With sequencing and bioinformatics, they found the genetic basis for this disease in a matter of hours.
- Yup, basically the right idea for polygenic risk scores. All of these approaches are only useful for diseases that can be actually predicted based on genetics—which obviously isn't everything. Lifestyle, diet, etc. are still enormously important, but PRS scores and clinical sequencing are super important describing the portion that is explained by genetics.
- the cancer example is actually detecting DNA that tumors leak into the bloodstream. While this approach is primarily used for cancer right now, there are definitely people and companies working on predicting lots of diseases (and even broad physiological scores of aging) from blood samples. Definitely compelling for auto-immune diseases too.
- I'd argue that the results I linked to actually explains part of the difficulty of diet science. Until now, there has been an assumption that most people respond fairly similarly to the same food types. The paper linked shows that this is not the case. Some people have massive blood sugar spikes after eating brown rice but not after eating ice cream, and vice versa. Part of this is predicted by the gut microbiome (and genetics too). Without measuring this, it is very hard to actually know how food will impact you. In general, I think that CGMs are likely going to be used for healthy people and not just diabetics, so that we can figure out what different foods are doing to our blood sugar.
Look forward to the rest of this series Elliot! Especially curious to hear your thoughts on DNA synthesis, and how enzymatic approaches might evolve over time.
I might also add that beyond breakthrough technologies enabling us to understand how our genomes impact health, our ability to make meaning out of this new information is critically important. I shared some brief thoughts on this today: https://healthandwealth.substack.com/p/genetic-counseling
Thanks for sharing Christina, this is a good point! I'll have to check this post out.
Great post! Given the general interest focus of this series, I hope you won't mind some noob clarification questions (coming from someone whose last exposure to Biology was AP Bio in the early 00's) about the
"[Sequencing data] will have uncontroversial benefits like finding better drug targets for diseases"
- Can you provide an example?
- If we've already sequenced a bunch of people, can you give some intuition around how and why sequencing many more will lead to finding better drugs for certain diseases?
"[Sequencing data] will greatly improve precision medicine-where patients are matched with drugs based on their genetics"
- Can you give an example?
- How can we learn this relationship? Even when the health outcomes are easy to measure (not the case in mental health, for instance) how exactly would we be able to detect that certain drugs are more effective for people certain genes? From a statistical standpoint, if there are dozens of genetic subpopulations that may respond differently to dozens drugs for a given disease, doesn't that lead to an explosion of treatment groups during clinical trials?
"Sequencing will also be used more directly in clinical care," and the two examples you give are diagnostic.
- Can you explain the first example? (What's a causal mutation?)
- In the simplest case, I understand how cheap sequencing could detect things like congenital single gene diseases. And though I don't understand what polygenic risk scores measure, exactly, I understand being able to say "people with a genetic makeup similar to yours go on to develop [insert disease here] x% of the time." And how that would lead you to take preventive steps to addresses [insert disease]. So, given that we already do seem to do this when people's lifestyle or demographics are similarly predictive of [insert disease], how much extra predictive power does sequencing give us?
- Going beyond "you were born with these genes, and we know that causes or is at least a risk factor for [insert disease]," the cancer diagnosis example seems to imply "we can detect disease before it fully develops because it's changed your DNA." Are there any other diseases that we can detect in this "non-congenital" way?
"[Sequencing gut microbes] could become essential for understanding how different diets actually impact our bodies on an individual basis." The dismal state of nutrition science suggests that we already have a hard time measuring the relationship between diet and impact on our bodies (likely because there are so many input variables, and the interesting outcomes take years if not decades to manifest). Would adding the additional variable of gut micro composition somehow make this topic easier to study, instead of harder?
Thank you!
Awesome, thanks for reading! Great questions. Some answers:
- On average, drug targets that are supported by statistical patterns discovered in human genetics are more likely to succeed. Based on this, there has been a big shift in biopharma to use more genetics in drug discovery. The UK Biobank project I mentioned is actually a public-private partnership with several pharmas using the findings for drug discovery. More on this here: https://doi.org/10.1038/s41588-021-00885-0.
- Sample size is really important for human genetics, because some important rare variation only happens at low frequencies in the population. Because ancestry is important, we also need to do a much better job of sequencing non-European populations. One of the preprints I linked to does a nice job illustrating this: https://doi.org/10.1101/2022.01.07.475305 as you sample more you start to saturate the total predictiveness of genetic information, but it is on a population-specific level (here only for Europeans).
- A person's genetics can impact how they will respond to a drug. This is true in cancer, where only certain mutations are treatable with tyrosine kinase inhibitors. It's more broadly true than cancer though, for example your HLA haplotype determines how well you will respond to certain antiretroviral drugs for HIV. There's actually a whole field working on this type of problem called pharmacogenomics, with many types of interesting stories in gene-specific drug responses: https://en.wikipedia.org/wiki/Pharmacogenomics
- The answer to your second question is linked with the first. Making these discoveries requires large genotyped drug response databases, and is still an active area of research. It typically involves finding how certain genotypes differentially metabolize drugs.
- A causal mutation is a sequence change in a gene that is causing the observed phenotype—in this case an inherited disease. So the key goal in clinical genetics is to identify this mutation to solve the case. In this case, it was an undiagnosed patient with an unexplained heart defect. With sequencing and bioinformatics, they found the genetic basis for this disease in a matter of hours.
- Yup, basically the right idea for polygenic risk scores. All of these approaches are only useful for diseases that can be actually predicted based on genetics—which obviously isn't everything. Lifestyle, diet, etc. are still enormously important, but PRS scores and clinical sequencing are super important describing the portion that is explained by genetics.
- the cancer example is actually detecting DNA that tumors leak into the bloodstream. While this approach is primarily used for cancer right now, there are definitely people and companies working on predicting lots of diseases (and even broad physiological scores of aging) from blood samples. Definitely compelling for auto-immune diseases too.
- I'd argue that the results I linked to actually explains part of the difficulty of diet science. Until now, there has been an assumption that most people respond fairly similarly to the same food types. The paper linked shows that this is not the case. Some people have massive blood sugar spikes after eating brown rice but not after eating ice cream, and vice versa. Part of this is predicted by the gut microbiome (and genetics too). Without measuring this, it is very hard to actually know how food will impact you. In general, I think that CGMs are likely going to be used for healthy people and not just diabetics, so that we can figure out what different foods are doing to our blood sugar.
Hope these answers are useful!