Machine learning/statistics vs. computational biology for Masters/PhD?

[ijop45]

I'm struggling with making decisions about (and determining my chances for) grad school. I'm an American graduating from a Canadian university, hoping to apply to American/European Masters & PhD programs.

I'm graduating soon from University of Toronto with a 3.25/4.00 CGPA (low mostly due to a terrible first year), research experience (1.5 yrs; hopefully 1-2 publications accepted in quality journals by the time I apply), and good recommendations from both my PI (who is well known in his field and in bioinformatics as a whole) and professors. I specialized in computational biology and majored in neuroscience. My research has been a mix of applied statistics (genomics) and applied data science (mainly with clinical relevance). I plan to take the GRE and expect to do well based on past performance on standardized tests. My coursework has been fairly evenly split between CS (5/20 credits) and biology (5/20 credits), with a bit of math (3/20 credits), stats (1/20 credits) and a few physics/humanities credits.

I'm hoping I can get some opinionated answers to a few questions about the process from people who have applied to ML and/or computational biology programs.

1. How is your Canadian (esp. UofT) degree/GPA/undergrad coursework viewed by schools in the states/Europe?

2. If you worked in both applied statistics/algorithm development and applied data analysis, which did you prefer? What was your research like once you chose a path? Did you always have an aptitude for algorithm development, or did it come with maturity & familiarity with the types of data common in your subfield of biology?

3. If you went into applied statistics/algorithm development, what was your theoretical background like? My interests lie in networks/clustering/unsupervised techniques, and I worry that my math/stats background isn't sufficient (only 1.5 courses!)?

4. Is there a profile for admits of stats Masters/PhD programs at schools of the same caliber of UCLA/UCSC/UCSF/Georgia Tech (can you tell I'm sick of Toronto winters?) for bioinformatics or applied statistics/ML? There don't seem to be many relevant threads on TGC. I'd like to know if I have any chance at all, because I know my GPA is kind of shit and my coursework is not ideal for stats.

5. I'm considering not applying to grad programs this fall, working in a relevant position (maybe research-oriented) and taking a few extra courses with a focus on math/stats, getting my GPA up to 3.4-3.5, and applying next year instead. Has anyone applied to a program, gotten rejected, and re-applied after bulking their application the following year?

I've looked at several schools, and I find myself drawn toward the statistics/ML researchers because their interests align with mine. But I'm not convinced that I have the background to apply for a Masters or PhD focused on statistics (ex: at UCLA, there's a 20% admit rate for the stats Masters vs. 24% for bioinformatics PhD, with an order of magnitude more applicants for stats). My interests still lie in applications to biology over, say, computer vision or any other field, but I've little interest in spending the majority of my time applying basic supervised approaches to straightforward biology problems, as one likely would if they worked in a biology wetlab as a data scientist.

Any thoughts at all would be appreciated.

[Quickmick]

Hello ijop45,

Just curious, what do you want to do after you get your degree? Thinking of the endgame might help you plot a course. Also, when you say 

On 6/30/2018 at 11:09 PM, ijop45 said:

I've looked at several schools, and I find myself drawn toward the statistics/ML researchers because their interests align with mine. 

you might consider looking at the researchers first, then where they are based.Maybe look at your most cited researchers? Someone who inspired you somehow? One thing is for sure, you won't get into 100% of the programs you don't apply to! Hope this helps a little, just thought it might be beneficial to re-frame how you are looking at potential targets, then when you have some work on the strategy to get there. 

I do a lot of predictive work using ML. For me, it is just the best tool for the job...are you interested in writing the tools or using the tools? 

A colleague of mine (in the US) has a degree from a Canadian university... how it is 'viewed' has never come up.

 

good luck!