To all budding compbio & ML folks interested in bio: Don't just only run behind the latest ML model hype train. The greatest long run impact will come by really assimilating prior bio/compbio literature with the goal of really understanding strategies for how to model biology. 1/
8
129
784
Replies
And by collaborating closely with biologists developing technology, deeply studying biological systems & diseases to identify the relevant questions in biology & medicine. 2/
1
1
69
This is not to say that u shud not keep pace with the latest ML tech. The tech is amazing & has immense potential. But to make impact in bio, in most cases, u need to understand how to formulate bio problems as ML tasks, 3/
2
5
66
whether the data u have is appropriate for the question & the modeling, what kind of data you might need to answer a question or a family of questions 4/
1
1
37
and understand deeply the strengths & weaknesses of different ML models in the biological domain (not just in other unrelated domains) 5/
1
1
29
And I can guarantee you, you do not need ultra large models and insane compute to address a large majority of extremely exciting problems in biology via computational approaches. 6/
3
4
71
Also please respect the data by delving deeply into how it was generated & processed. Be humble & respect the immense body of prior work. But there are plenty of unsolved problems that require deep ML expertise coupled with a willingness to truly embrace biology. 7/7
3
8
93
@anshulkundaje
Given what you know now, How would you start learning ML and DL if you had to start all over again? What would you learn and in what order
0
0
10
@anshulkundaje
Yeah exactly. It really feels like some ML folks see the latest trend as a solution in search of a problem. Tailored solutions to biology requires some understanding of it and collaboration with domain experts who know what’s already been tried before.
0
0
6
@anshulkundaje
do you have any suggestions for people who have a biological background and want to be in compbio?
1
0
0
@anshulkundaje
@StanfordAILab
How about a nice reading list that is curated and takes you from noob to cutting edge?
0
0
1
@anshulkundaje
Totally agree! For example, Bayesian approaches have a lot of potential, but I always struggled to find great resources to learn that
0
0
0