My postdoc @JosephMatheson2 and I were sending off my former undergraduate student Minkyu Kim to graduate school at @USC and giving him some advice over a beer. I thought it might be useful to share.
1. Grad school is a marathon, not a sprint. Focus on finishing rather than on running fast.
2. Get comfortable with not knowing things.
3. Grad school is by far the best time in your life to learn. In learning, go both deep (become an expert in something) and broad (learn beyond your immediate subject of study). To learn, read, go to seminars, talk to your fellow grad students and professors, go to conferences.
4. There are many skills to learn, e.g., communicating your work, writing, time management, etc. But the most difficult skill to learn is to identify problems that are interesting, i.e., worthwhile solving, and solvable (by you!). Be deliberate about honing this and other skills.
5. Establish professional relationships. Overcome your shyness and insecurities and talk to fellow grad students and to professors about your research. These relationships will likely help in your future career in ways you can’t anticipate.
6. The relationship with your PhD advisor is probably the most important professional relationship that you will have. Be deliberate about developing it. Give your advisor feedback on what works for you and what doesn’t.
7. If you feel you are in some sort of trouble, seek help. Don’t let problems linger.
What would you add?
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I’ve been thinking recently about the generality of scientific results. Lack of generality is a common and often lazy critique of papers and grants. But how do we actually know whether a result is or would be general? And what is “general” anyway? 1/26
Here are some simple observations that helped me think a bit more clearly about this topic. The first most basic observation is that generality pertains only to mental constructs (theories or models), not data. Data are data. The interpretation of data can be general or not. 2/26
Interpreting data is of course necessary for understanding. So, we need theories and models. In fact, one of the main goals of science is to produce theories that describe as much of reality as concisely as possible. (Somebody else said this, I am paraphrasing.) 3/26
@GrantKinsler, @KSamerotte, @PetrovADmitri, I read your paper over the weekend. So much to think about. I certainly haven't fully digested it, but it’s amazing work!
We started thinking along similar lines in our paper by @erjerison et al. nature.com/articles/s4155…
Your “fitness profiles” are our “pleiotropic profiles”. We also did dim. reduction and saw mutants form clusters. But you pushed this approach so much further. Really cool!
The key new insight that I learned from your work is that a set of mutants will live in a space whose dimensionality is the number of independent effective phenotypes that they perturb.