So I just published the Longevity FAQ. This represents my first blogging foray into molecular biology. So it happens, I hadn't really studied biology since high school, so here's how -and why- I went from zero to writing a FAQ on a complex topic.

Why write about it, in the first place? An original motivation is that I've said a bunch of times that I see more progress in the future coming from the life sciences rather than physics, yet my knowledge in biology was very, very lacking, not having ever seriously looked into it. On the other hand ageing had been a topic in the background for me for years; back in 2009 or so I had read Aubrey de Grey's book Ending Aging. More recently, and wanting to learn more about it, I noticed that there was a lack of something like what eventually became the FAQ, something that explains, from first principles (At least as far as the functioning of the cell is concerned) what is going on, explains core concepts and their relation, points to contradictory evidence, the different changes in views over time, and nuanced results, plus therapies.

So to start getting that background knowledge started reading the textbook on Molecular Biology, the Molecular Biology of the Cell. As much as I'm more of a fan of papers as being more efficient media for the transmission of knowledge than books; textbooks in particular are the best way to get into a new field. A textbook has a structured approach starting from the very basics, and it will also tell you things you didn't know you wanted to know in the first place.

But lacking super-basic knowledge of biology, I had to resort too to the Khan Academy biology courses plus assorted resources online to complement the textbook. I didn't work through the whole textbook, of course, only sections that I thought were relevant. This left me with some knowledge to understand topics I would be looking at, but at the same time I ended up with gaps in very basic knowledge (e.g. I wasn't aware of a key process that underlies the adaptive immune system)

David Sinclair's book was the immediate trigger for me writing the FAQ, I had read the book and ended up wary of some conclusions, he seemed very dismissive of the role of ROS and DNA damage, in favor of his own theory that puts sirtuins at epigenetics at the core, To some extent my concerns were valid: He was indeed downplaying the role of the former. My intent, to be clear here, was not to fact-check the book; overall I was not skeptical of the main argument. The book already mentioned many of the key protagonists in aging story: telomeres, mitochondria, reprogramming, sirtuins, senescence, and more, and it linked to many surprising results plus foundational work like the Hallmarks paper, so I initially used that to inform my information gathering.

Usually when I write about topics, I just search for the information I need in the form of literature reviews or meta-analysis, looking at recent work and writing in the blogpost directly.

The first version of the FAQ started like this, but halfway through it I decided to try a different approach. I knew that my usual way of going about writing has led in the past to missing some studies, and in general favouring a subpar style and flow of the argument; as I don't know what I will be writing about until I've gotten to the relevant section. Instead, what I did here was to delete most of the original document, and decided to use a Notion spreadsheet (Though the Notioneers may object to Notion being described in such way) to store the papers I was reading, tagging them with the topics they covered and the year of publication. That way I could take notes with the core results from each paper, and be able to easily filter them by area and sort them by year, enabling me to see how the literature evolved over time; generally all else equal more recent work should be more correct, as the extra time may have allowed for more arguments to be considered and more evidence to come through.

I was aided in this quest by a tool I developed, which has a database of most scientific papers(Over 150 million), While I haven't made public all the API routes and features yet, I was able to do queries like "Find all papers that cite X and Y" which is useful if papers X and Y contradict each other and one wants to find a paper that synthesises them both, or otherwise explains what the disagreement is about.

So at some point I sat down and wrote a structure for the article; first I laid down some basic concepts, with the relevant caveats and studies linked, the sort of thing I would have liked to have when I started, then I explained core concepts; I picked up these concepts because they tended to appear over and over in the papers I was reading.

Then, I had a therapies section; here I was largely inspired by Laura Deming's FAQ and the papers that looked at precisely anti-aging therapies in isolation; I decided to take the union of those sets of therapies, discarding some that seemed less relevant.

In looking for studies, when finding that something seemed surprisingly good, as is the case for senolytics or rapamycin, I tried searching for terms like "rapamycin ineffective" or "senolytics negative effects"; the latter led to some interesting speculation as to how that may be. In general I applied the standard Nintil method:

  1. Look at literature reviews on the topic
  2. Expand and search about topics you don't understand
  3. For results you are skeptical about, or that different reviews contradict each other, find literature reviews for those concrete points.
  4. Follow the citation graph until the most recent evidence has been reviewed
  5. In the process, take note of how much disagreement there is, the journals papers tend to appear on, what are some papers everyone cites, etc.

For an individual paper, there's the question of what to read. At first, with the very general literature reviews, the answer is everything. In the more narrow literature reviews, I admit I'll skip sections and focus on the figures, introduction, and conclusion; usually these sometimes get too deep into a particular topic and it wouldn't help me that much (It won't make it into the FAQ and I will forget the details anyway). Sometimes I would have concrete questions; in a section of the FAQ there is a brief reference to the % of senescent cells in adults and young individuals. This is something I had wondered during writing the FAQ, as it's a common assertion that "just a few senescent cells can have large effects", but how much few is that few? So I looked up papers for that, and in that case I just took the number and read the paragraph where it appeared.

A problem that occurs when attempting to reconcile disagreements in the literature is who to trust. Sometimes the disagreeing paper came later and then even later one can find a joint paper by both authors where they give an interpretation to the results (As was the case with the monkey studies). Another case is that a later paper says the earlier paper was flawed due to poor methods, or that the earlier paper claimed results that were not generalizable. This was the case for the Whitaker (2013) paper I cite on the effect of sirtuins in flies. If the later paper references earlier work and explains the disagreement, I'm prima facie willing to accept that as the truth.

But there's an issue where one finds what seems to be different communities of researchers (This might be the case with the free radicals researchers) who tend to publish on different journals and accept different things as true, odd as this may seem. Of course, here the problem may be that one side has not read the other's evidence, or that one side's arguments are just not good. This was the case in the ROS section. Here, the best I could do is do a longer section, noting down the results people have claimed over time in reviews; but I cannot have a lot of confidence in this section; to do so I'd have to dig deeper and understand better the different assays and pieces of evidence. So if I did a deeper review of the role of ROS I'd begin by noting down specific claims (e.g. antioxidant intake does not lengthen lifespan, in any animal model, superoxide dismutase knockout reduces lifespan), noting down the evidence given and one by one check the discussion for each fact. Then once I have individual pieces of information, I could see how they could logically fit together and determine who is right.

After publication, as I knew I was bound to have many mistakes, this being a topic new to me, I made the commitment to update every day the FAQ with whatever corrections I have sourced that day from feedback I'm receiving from readers. On day 2 after publication I also decided to send the FAQ to various reputed ageing researchers, to see if they have any corrections to ad (As of the writing of this, they haven't replied).

Early response seemed good, a biogerontologist on reddit said it looks pretty good (8-9) out of 10, likewise I got some very good responses on twitter. The changelog in the post reveals that there were a number of mistakes in the original that I've corrected. Doing a post-mortem, most of the serious mistakes were to be found in sections of the FAQ that were in the initial version, which I started to write before a comprehensive reading of the literature. The obvious fix here is to wait until everything has been reviewed to start writing. The second class of mistakes was due to lacking basic knowledge about biology. This was to be expected.

At the end of the day, I could have chosen to be more careful and spend more time reading about the basics before deciding to write a FAQ on a complex topic I had little idea about prior to starting. But instead, I judged more efficient to try anyway and then use the feedback I would receive as guide to know what I was missing. As a result, I have identified weaknesses in my knowledge of the immune system, the mechanics of cancer, atherosclerosis, and inflammation; which might make it into an updated version of the FAQ.

I am fairly satisfied with the final result; I predicted that it would be a reasonably good FAQ that would contain some mistakes, and that it would get polished over time, and that has been the case.