Biology Meets Programming: Bioinformatics for Beginners

Of the half dozen or more MOOCs I've taken, this class is in the top two. MIT 7.00x, taught by Eric Lander, is the only one that is on the same level. This is the highest compliment that I can give; I loved 7.00x.

The prerequisites I recommend are an introductory DNA class (MIT 7.00x or equivalent), some programming experience, and (optionally) some formal exposure to algorithms. I do not think someone whose lifetime programming experience consists of completing a single introductory Python class will do well, though it is possible. You are allowed to complete the assignments in the language of your choice; I chose C++.

The material is presented as a collection of video lectures, and an "interactive text" that breaks the written material into many small, easy-to-digest pieces. I typically watched a video and then immediately dived into the corresponding text. On the rare occasion that I had trouble with the material, the discussion forums were a good place to get help.

You are graded on homework (programming assignments) and quizzes. The quizzes are easy if you read the material and do the assignments. The "heavy lifting" is the assignments, not so much due to their difficulty as to their quantity. There were 55 graded assignments, and at least a dozen optional assignments, some of which taught you how to approach the graded assignments. Each assignment asks you to perform some computational task on some data. You are provided with a tiny dataset which is useful for program development, and a practice dataset which is usually a good predictor of whether your program can handle the graded dataset. When you feel you are ready to be graded, you are given a unique dataset and 5 minutes to return the answer. If you fail to give the correct answer, you can try again, as many times as needed.

If this sounds like a lot of work, it is! But don't be scared of it. This class reaffirmed for me the truth of the saying, "you get out of it what you put into it". I got a whole lot out of it.

I give this course the full amount of stars, and I really mean it. This course was amazing, among the 45+ MOOCs I have taken it was one of the most engaging and nicely delivered when it comes to the interactive text!

What you gain from this course is perhaps 20 percent of biology and genetics (if you are new to the subject but highly interested, you really get something out of it), and 80 percent computer science and algorithms. The latter is where the meat is in this course, and it does it, if I judge by my own experience, a tad more effective even than Design and Analysis of Algorithm by T. Roughgarden. However, I may be biased, because, probably if you have the two Algorithm classes of Standford under your belt and get the different perspective this MOOC takes on those concepts, you probably enjoy the course even more, being already familiar with some algorithms.

The programming assignments are ample in number. Their level of difficulty should be doable for people with some CS background, and those who have implemented some basic algorithms before. A drawback is (as naturally in a computer-graded course) that you 'only' have to submit the correct resulting datasets to thhe grader -in at most 5 minutes time (which, OK, already puts some demands on implementing not terribly slow algorithms) - but the actual efficiency and elegance of one's programs is not assessed. However, that's is not a major weak points, because it remains the participant's own responsibility to produce decent code.

I did a mixture of strategies: sometime watching the videos before reading the interactive text, sometimes only reading the text, and sometimes reading before watching; the most effective for me was the latter. But you can actually study without the lecture videos (it works, and the staff's proclaimed intentions included figuring it out whether is does the trick). It does!

A tip for the choice of language: I have used three languages for the assignments: R (which turned out to be too slow for the data-intensive problems), C++, and Python. The most efficient choice in terms of implementation time was Python, naturally, and it was good enough in terms of running time even for the fleshy tasks. If you have basic knowlegde in Python, it is a time-saving choice (lists and dictionaries were sufficient for all the situations in which you needed data structures) if you do not have the motivation to spend extra time on writing C++ code.