As for me, I would recommend a broader range of topics than one or two Open Source projects. It depends on the students. Some may be more literate in biology than computers, and visa versa. And, some may need an overview of the many basic problems in bioinformatics. A single Open Source development project usually focuses very narrowly on a single problem. But, if your students have the basics down, it may be a good excercise.

I believe in learn by doing, that actually is how I went through from biologist to web developer/consultant. Some of the basic programming skill learnt from class will make your life a lot easier and you don't have to learn by mistake. but there is some advantage of not take formal train class also, like you will have a different approach to problems, you will have broader view to the field than just learn from some class. since we are in such a new field, you never know who got the right solutions to problems. So I Believe get into some interesting and key projects in the field and try to solve some important problems will be a great way to learn. At least it is as good as taking classes, if not better

It is really great that you guys can get together to do something like this! It can be really hard to get into bioinformatics when there isn't an established program at your university. In terms of the class vs. open-source debate, why not try to get the best of both worlds? Many people need a structure class environment to keep themselves focused (or else, why would we be going to grad school :-), and open source projects are quite unstructured by their nature. On the other hand, it is nice to be able to do work for a class that feels like it matters, which would make contributing to an open source project a good idea. Of course, doing both is probably a big challenge, but on the other hand it helps to have some of the fundamentals o' bioinformatics stuff down before you feel conformatable contributing to a project. It is also a challenge to find people who like to program enough to want to contribute to an open source project. But it would be super if you could manage to get it all together, there are plenty of great projects out there that could use help :-)

I am a teaching assistant for the Canadian Genetic Diseases Network national workshop series in Bioinformatics: http://www.bioinformatics.ca I'm also involved in open source software development, with Piper: http://bioinformatics.org/piper I think that bioinformatics course-work is very good at providing the conceptual framework necessary to understand bioinformatics theory, but it is largely lacking in providing the computing skills, which come with time and practice. Participating in open source bioinformatics software development really does a lot to impart the practical skills necessary to become a competent bioinformatics scientist. So I advocate both. I think it would be interesting to design an online course that includes developing open source bioinformatics software as part of the 'practical component'. Distributions such as bioperl, biopython, bioxml etc., could be used to provide examples of how to code good bioinformatics software. I would be interesting in helping to create something like this if there were enough interest to make it worthwhile.

I am teaching a graduate course in Bioinformatics at UMassMedical School with a collegue from Boston University. He teaches a Bioinformatics course at BU also. We cater to biologists who want to use bioinformatics, but for the most part do not have a desire to program or to develop new algorithms (math/statistics). We started with a simple problem, comparing two sequences, and through a tools approach, progress to more complicated problems, multiple alignments, database comparisons, etc. In the end students are using very sophisticated techniques. What underlies all of this is a statistical approach to making comparisons between different pieces of data. That said, what is your focus? There are a lot of topics in bioinformatics and each runs the whole spectrum of skills , biology-programming-analytical. In the end, bioinformatics is just the traditional informatics approach ( gather, store, retrieve, process data) applied to biology data. I think the new thing is discovery, using bioinformatics to discover new things (data->information->knowledge). Oh, I forgot to talk about Open Source. It's great. Don't use anything else.

To answer more directly. I would indeed form an interdisciplinary group. I would make sure you included some stat-math types. A close cousin of bioinformatics is computational biology. There is a difference.