This may be a tad late, but as someone about 6 weeks off from defending my MSc in Bioinformatics (and having watched a couple of people from my cohort complete), I may be able to give you an idea of what I've seen at my institution (a large multi-hospital research institution in the US).
What are the expectations for MSc students in bioinformatics?
I'm not sure I'll be able to frame the requirements as you have. That said, for the MS, our guidelines state that our thesis project can consist of, and I'm paraphrasing here: a novel implementation of software to an existing problem (i.e. your software improves on existing solutions in some way), a significant improvement to an existing software, or the application of current software to a novel problem. Fortunately, being a large clinical institution, we have no shortage of novel problems to solve.
So, what does that mean regarding the breadth of the project?
- For an MS, we often don't generate our own data but may instead take data that already exists from research done by a lab on campus, a lab at an institution we're collaborating with, or public data sets (e.g. TCGA, ENCODE, ClinVar, Reactome Pathway Database, Uniprot, etc.).
- We'll then mine this data to get at our biological problem of interest, use it as a development set for an algorithm or method, etc. Another approach may be to generate a novel data set by combining multiple data sets in a clever way.
- Keep in mind, there are a number of highly respected bioinformaticians, Atul Butte at UCSF comes to mind, who have launched successful companies by mining public datasets and combining their domain knowledge and experience – there's a lot of public data out there waiting to be transformed into knowledge.
For masters, is it possible to take an open source tool and build on top of it to solve some issue it has?
As you may have guessed by now, I'd venture a yes – provided your improvements enhance its value in a significant way (particularly in terms of reproducibility and validity, but usability matters too) and is already a widely used tool for a specific problem or (ideally) class of problem. That said, whatever improvements you make should target a specific problem and improvements should be measurable in some way.
Regarding figure counts, I wouldn't worry too much about them. For an MSc, they're less relevant given that you probably won't be producing your own data rather than conducting a deep analysis. I've generated hundreds of graphs, tables, etc. throughout the course of my research in an effort to understand the underlying complexity of the data and the problem at hand. Visualization is a necessary and powerful tool. I'll likely include somewhere between 8 and 12 in my final thesis writeup.
Hope this helps.