Hello folks,
I have studied computer science and ML for a while now, and am starting to get in the amazing world of Bioinformatics, I would like to know if you have some text, tutorial or suggestion on where to start or what kind of open problems are there in the area.
Thanks a lot
Leon
Machine learning has certainly found wide application in bioinformatics. A trivial example: search PubMed for the phrase 'support vector machine'. There are currently 2262 results applied to diverse problems such as predicting protein-protein interaction, identifying features in nucleic acid sequences, analysis of microscopy images and the physiology of muscles during exercise.
A word of caution regarding "big, unsolved problems." Bioinformatics is not really like this: we are not looking for the Higgs boson, or figuring out what came before the big bang. The vast majority of day-to-day bioinformatics consists simply of helping biologists to get more from their data. Very often, this requires nothing more than the intelligent application of existing tools from mathematics, statistics and computer science. You should realise that for many biologists, the automation of a manual procedure taking many hours using a simple shell script that takes milliseconds is an absolute revelation.
The key thing then, when entering the field, is to figure out the problems faced by biologists, not the theoretical problems that take your fancy. And the best way to do that is talk to them. That way, you can identify those areas where you can best apply your expertise and bring something new and useful to the table.
There is an excellent review on the topic, presenting both methods and applications:
Machine learning in bioinformatics, by Pedro Larrañaga.
Check out his website for other interesting publications.
Vasant Honavar, one of the professors in our computer science department, specializes in machine learning and has extensive collaborations with biologists on campus. If I wanted to get a general idea of machine learning's applications in biology, I would take a look at a list of his publications.
Turns out his homepage has a link to such a list, and it looks like he keeps it up to date pretty well!
Our group (Rost) is doing a lot of machine learning in the field of structural bioinformatics. We try to predict structural/functional features of proteins based on sequence alone. Check out our lab's homepage [1]. Chris
[1] www.rostlab.org
I recommend you to read "Pattern Recognition and Machine Learning" by Christopher M. Bishop. It is a good tutorial for bioinformatics or computer science student to learn ML for bioinformatics question, such as ANN or HMM for gene classification and gene prediction.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
IMHO, this clause "The vast majority of day-to-day bioinformatics consists simply of helping biologists to get more from their data" will be ideal only for bioinformatics service/core group, not for a bioinformatics research group.
Doesn't research help biologists get more from their data? I don't make these distinctions. If I help you get results quicker, we're both engaged in research activity.
still, I'd say there's plenty of big unsolved problems -luckily for us, young computational biologists ;). To name a few: ab initio protein structure prediction, remote homology detection, GWAS statistical treatment (detection of low frequency and low effect genetic variants)...
Oh yes, there are plenty of interesting problems and I'm not saying all bioinformatics is the humdrum, information management variety. I guess my message is: make sure your work is useful work. Remember, biologists don't read bioinformatics journals :-)