,
13 tweets,
3 min read
Read on Twitter
1. This week I have been thinking about how difficult it is for a lot of scientists to deal with subtlety in our data, results where the answer isn’t a binary yes no or are just plain messy. So a thread about subtlety in science!
2. When I was a new grad student, I read an essay by one of my scientific heroes Sydney Brenner. He describes that in his time a meaningful result was one that (I’m paraphrasing) you would see from the other side of the room even when graphed on a 7-cycle log graph paper.
3. I loved that approach, it invoked a world where things were clear, the data gave you a clean yes or no answers, there was no subtlety. But being inexperienced I didn’t think much of the fact that he was talking about bacterial & phage genetics in the late 50’s/early 60’s.
4. I do think the attitude expressed by Brenner, where the data always had to be clear-cut & conclusive was pervasive amongst generations of molecular biologists/geneticists including those who trained me, and this greatly shaped my worldview.
5. However, the more I started to work with actual complex living systems (as a developmental cell biologist) the more I realised that the view Brenner expressed where things had to be one thing or another doesn’t work very well in many contexts we scientists, encounter in nature
6. For example one of the mutants I worked with (rhea) caused a phenotype (failed dorsal closure) about ~50% of the time. I still don’t know why this is, 17 years later. You can’t show that on a 7-cycle log graph paper but that didn't make the result any less meaningful!
7. Particularly relevant to my own work I noted how many geneticists have this idea that wildtype has 100% function, a heterozygous also had ~100% function (which is why het/wt is so often used as a control), and whatever was in the database as a “null allele” had 0% function.
8. But this is not what I encountered, I kept encountering “null alleles” with a significant amount of function, or heterozygous mutants that made much less protein than the wildtype & showed defects, or phenotypes that were present in some of the mutant animals but not others.
9. I also realised that my colleagues from the fields of functional genomics and population genetics had a much more sophisticated view of what wildtype/heterozygous/”null allele” meant when compared to what functional geneticists like myself thought.
10. Also frustrating was that when I looked at published papers, data was often presented according to the model exemplified by Sydney Brenner, things were clean and clear, heterozygotes never had phenotypes, a “null allele” always showed the exact same phenotype.
11. This meant that when scientists try to introduce some complexity & subtlety into the stories they tell (already hard due to space restrictions at many journals) there is often lots of resistance from reviewers & editors, who are used to seeing everything neatly tied together.
12. However I see several developments in science that will make it much easier to introduce subtlety & complexity into our work. Things like the amazing quantitative tools we have, the rise of systems biology, advanced imaging approaches. These tools foster complex thinking.
13. But I also think we should make a conscious effort to be more open to data that is complex, messy, or that goes against our need for neatly packaged clear-cut stories. Our work as scientists should be a reflection of nature and not our need for straightforward answers.
