November 27, 2017

What have I been reading?

In The Berlin Project, Gregory Benford, who is deservedly known for writing some of the best hard science fiction around, asks what would've happened if the United States had developed a nuclear bomb early enough to use it against the Nazis during World War II. The book is very well researched, well written, plausible and makes you think. What more do you want from alternate history? Recommended.

Max Steele doesn't even have his own Wikipedia page. I feel that this is an injustice and that he should be more well known, but not strongly enough to actually do something about it and start that Wikipedia entry. Also, I don't know anything about him beyond what I can guess from reading two of his books, both of which I suspect are vaguely autobiographical. Debby plausibly describes what the inner life of a person with intellectual disability might look like. However I liked the short stories in the collection The Hat of My Mother even more, and it's one of the best books I've read all year. Hat tip to my mother for recommending this one to me.

Because of where and when they're set (1930s South), Steele's books reminded me of one of my all-time favorite novels: Look Homeward, Angel by Thomas Wolfe.

Mary Beard in SPQR provides an overview of the history of ancient Rome from its founding to the first century AD. She clearly knows her stuff and I learned a lot (for example I wasn't aware how much the Romans where sticklers for the rule of law), but I can't say I enjoyed the book because her prose is a little dry. If you're interested specifically in what ancient Rome can and can't teach us about what's going on with America, I highly recommend Vaclav Smil's Why America is Not a New Rome.

I picked up a copy of Vince Flynn's thriller Act of Treason that another passenger had left behind when deboarding a plane. It's a rather entertaining action novel starring the all-American CIA operative Mitch Rapp, giving some terrorists what he thinks they're deserving (death in most cases). The blurbs make it obvious who the intended audience is: Glenn Beck thinks it's "Captivating", Rush Limbaugh thinks it's "Just fabulous" and Bill O'Reilly thinks that "Every American should read this book". One good thing that came out of doing just that was that it made me avoid the Mitch Rapp movie that recently came out. Thanks Bill.

The central thesis of Iron John by Robert Bly is that men don't do themselves or society any favors by repressing their wilder side. Iron John contains some interesting ideas, but its central contradiction is that it massively, comically overthinks what it means to "be a man".

More reviews to come soon.

January 12, 2017

Which scientific concepts ought to be more well know?

The answers to John Brockman's annual question on Edge are always worth reading. This year, the question was which scientific concepts ought to be more well know. If you haven't done so already, I recommend that you take a look at what the more than 200 contributors thought here.

With all those excellent essays, what resonated with me most were a handful that emphasized the importance of humility, not only in science, but in societal discourse more generally.

For example, Barnaby Marsh writes:
You might not think of humility as a scientific concept, but the special brand of humility that is enshrined in scientific culture is deserving of special recognition for its unique heuristic transformative power ... Scientific humility is the key that opens a whole new possibility space -  a space where being unsure is the norm; where facts and logic are intertwined with imagination, intuition, and play
Oliver Scott Curry:
Fallibilism is the idea that we can never be 100% certain that we are right, and must therefore always be open to the possibility that we are wrong ... Fallibilism is also the guiding principle of free, open, liberal, secular societies
Nicholas G. Carr:
if our intellect is bounded, we can never know how much of existence lies beyond our grasp
And finally Sam Harris:
Our scientific, cultural, and moral progress is almost entirely the product of successful acts of persuasion. Therefore, an inability (or refusal) to reason honestly is a social problem. Indeed, to defy the logical expectations of others -to disregard the very standards of reasonableness that you demand of them - is a form of hostility.
Do read the whole thing though.

November 16, 2016

What should non-geneticists know about genetics?

There are things that geneticists hardly ever mention when talking to their non-geneticist collaborators, probably because they take them for granted. Stating them explicitly may be helpful for those of you who frequently work with geneticists. I'm assuming that you already know a little bit about genetic variants and association studies.
  1. When it comes to genetic association studies such as GWAS, correlation really is causation. If a genetic variant is associated with a trait, it or a variant close to it causes the trait, assuming it's not a spurious correlation. Because DNA is read-only, it's not possible that the trait causes the variant. This makes genetics different from e.g. gene expression analysis, where the arrow of causation can point both ways. A differentially expressed gene can cause a disease, but the disease can also cause genes to be up- or downregulated. Typically, you have to do follow-up experiments to determine what's going on. Not so in genetics.
  2. Genotyping isn't sequencing. Frequently, people will say something like, "we have sequenced those samples" when they were really genotyped on a chip. The difference is that genotyping chips are cheap ($200 or less per sample) and typically produce data on several hundreds of thousands of known genetic variants. Sequencing is more expensive (more than $1,000) and produces data on almost all the variants in the genome, including those that haven't been observed before. Unlike genotyping chips, sequencing also delivers data on structural variants such as insertions, deletions and copy number variation.
  3. Knowing the causal variant isn't the same than knowing the causal gene. Most of the human genome isn't coding for genes, and it's not clear what it actually does, or if it does anything important at all. The majority of variants that have been associated with traits and diseases are not located in the coding parts of genes either. For those variants it's difficult to tell how they exert their effect. Some that are known to change gene expression are called expression quantitative trait loci or eQTLs. For those variants that aren't eQTLs, people often assume that one of the genes that are encoded in their vicinity is the causal one.
  4. Knowing the gene isn't knowing the effect direction. Even if you know through which gene a variant exerts its effect, you still don't know in which direction the effect goes. Take the example of a genetic variant that has two alleles, G and T. Assume the G allele is the risk allele for a disease, and it's located in the intron of a gene. This does not immediately tell you if decreased gene function is associated with higher or lower disease risk. Again, eQTLs come to the rescue, as they will tell you if the risk allele is associated with higher or lower gene expression, which are reasonable proxies for increased and decreased gene function, respectively.
  5. Genotypes are discrete, phenotypes often aren't. A genetic variant typically has several genotypes. The example variant from the previous paragraph with the two alleles G and T will, in a diploid organism like humans, have three genotypes: G/G, G/T and T/T. It may therefore be tempting to assume that genetic variants are great biomarkers, as they will unambiguously show if the trait associated with the variant is present or not, maybe with heterozygotes being something in between. Unfortunately, this is often not the case, especially for complex diseases that have many variants associated with them. Each of these variants contributes to disease risk only a little bit, and as a result, individual variants aren't very informative.

There's more, but this post is already too long, so I'll save it for another time.

November 10, 2016

Is a third party candidate the solution?

This election was ugly, and the next four years are going to be traumatic. But America will get through them.

Even if Clinton had won, the election would have left scars. Just as Trump is strongly opposed by a portion of the country, she would have been hated by a different portion. True, that portion would not have had a rational reason to be as terrified as many Democrats are now, but that doesn't mean that their views don't count.

Repeating a campaign like this in four years wouldn't do anyone any favors and contribute nothing to healing the divides. But for any campaign that involves a Democrat running against Trump in 2020, this is what would happen.

Trump has said that he's going to erase the achievements of the Obama administration, and half the country already loathes him. I can't see how a Democrat replacing Trump in four years wouldn't be loathed by his supporters in turn. A large part of the country hating their president, no matter their motivation, and each president trying to erase what the one before them has done, cannot be a healthy state of affairs.

But what if someone who is not particularly objectionable to either camp were to run as an independent candidate? They would get the support of Republicans who stood up to Trump, of moderate Democrats, and of course of independent voters. And most importantly, they would break out of the self-reinforcing cycle of partisan hatred that casts its shadow over America.

September 16, 2014

Is genomics past its peak?

In his excellent blog, Robert Plenge recently asked how far along the hype cycle we are with regards to applying genomics to drug discovery.

The absolute number of genomics papers published in 2014 is likely to be higher than in any previous year. But when normalising by the total number of biomedical papers listed on Pubmed, a search engine for biomedical publications, it seems like we may be past the Peak of Inflated Expectations. The proportion of papers that contain genomics in their title or abstract has peaked in 2012.


I found the number of publications each year by searching for genomics on Pubmed. The number of papers for 2014 is an extrapolation based on the number of papers to date.


August 18, 2014

Should every drug have to pass a clinical trial?

If a pharmaceutical company wants to sell a drug, it first has to prove that it works and is safe by doing a double-blind controlled trial. Everything else is quackery.

At least that is what I used to think, but now I am not so sure any more. A major reason for my new-found uncertainty has been Peter W. Huber's book The Cure in the Code. Huber is a senior fellow of the Manhattan Institute, a free market think tank. Whilst his book is political, it raises points that should transcend the political divide. His central aim is to show that the top-down drug licensing process, as it is currently implemented in the United States and other developed countries, is not in the long-term interest of patients.

Consider the rare diseases that we study at the Sanger Institute, where I am a postdoc. In many cases, they are caused by a single mutation disrupting a single gene. Because the mutation is rare, not enough people have the disease to make it profitable for anyone to invest in developing a drug.

Nevertheless, sometimes doctors are able to repurpose a drug that has originally been developed for a different disease. For example, if the causal mutation disrupts a gene in a particular biochemical pathway, the doctor may know of a drug that upregulates the expression of another gene in the same pathway, therefore compensating for the mutated gene. That is great when there is a drug that can be repurposed, which in most cases there is not.


This issue is not limited to rare disorders. Complex diseases such as autism or diabetes have a large genetic component, but rather than a single causal variant there are tens or even hundreds. Each patient has a different combination of variants, and it is therefore unlikely that there will be a single drug that works for everyone. This makes a personalised medicine approach necessary, where the drugs that are prescribed are adjusted based on the patient's genotype. As with rare diseases, this means that patients with less common genotypes may lose out because there are not enough of them to warrant the development of a new drug, and because no drug that has been licensed for another disease is likely to help.

The higher the level of proof that is required before a drug can be sold, the lower the number of drugs that pass this threshold will be. Always requiring the highest level of proof may seem like the safest option, but is not necessarily in the public's interest if it means that many drugs will ultimately be unavailable. There is legislation such as the US Orphan drug act and a number of accelerated approval rules, but they only partly address the problem. There are still plenty of examples of drugs that may work in a subset of patients, or that may in the future be repurposed, but that are killed off in the licencing process and are therefore not available to anyone.

What is to be done? Creating legislation that allows the use of any compound whatsoever for the purpose of treating people with rare diseases or rare genetic variants could be one option. Of course, such a laissez-faire approach to drug licencing would be controversial and could encourage unscrupulous practices.

Right now, it is unclear to me what the balance between the two extremes of complete permissiveness and complete top-down control of drug licensing should be, but I doubt it is the status quo. I would welcome any views readers of this post may have.

August 3, 2014

How important is health care system quality?

A few weeks ago, I came across an article comparing the health care systems of eleven developed countries. This made me wonder how much health care system quality impacts physical wellbeing, for which I used life expectancy as a proxy. I was expecting a strong correlation between the two, but in fact health care system quality doesn't seem to be that important.

Healthcare system ranking (from best to worst) and life expectancy of eleven Western countries. The correlation (r) is -0.32 (solid line), or 0.06 when excluding the United States as an outlier (dotted line). In either case, the 95% confidence interval of the correlation spans zero, meaning that the data does not support any relationship between the two variables
Health warning: Do not overinterpret such a small dataset. This is me messing around with a few numbers I found on the internet. I am not a health economist.