>>13553081>I thought maybe I should've just learned to use existing tools rather than working on the string matching algorithms and stuff like that.You're right that in practice, you're going to be using existing tools rather than brewing up whatever low level implementation from scratch. The point of Rosalind for computer scientists is to understand what biological process you're modeling. The existing bioinformatic tools are trivial to use. Interpreting their outputs is another story.
> An ideal project would be to try training a neural net with genotypes and phenotypes, but I get the impression that sort of data is not easy to come by. Evolution is also cool. Not super familiar with the other topics listed other than some friends who did biophysics telling me that structural prediction is a big open problem.If you're interested in this sort of work, there's lots accessible, low-barrier-of-entry type work to be done here. Global phenomics won't be solved in our (or our grandchildren's) lifetimes, but prediction of biochemical function from AA sequences is very doable. Use of non-alignment based methods ('machine learning') is one the most active areas of research in bioinformatics.
If this tickles your fancy, try using the recently released AlphaFold codebase to beat DeepNOG's GO accuracy. Deepmind did a great job in experimenting with feature representation and DeepNOG's features are very crude in comparison.
https://academic.oup.com/bioinformatics/article/36/22-23/5304/6050698Good luck, anon.
