but the whole system is geared to "publish or perish". Already thousands of scientists leave the field every year because they haven't produced sufficient publication churn to carry on working. Pubmed is a cesspit of junk, growing by tens of thousands of publications a day. In this hyper-competitive numbers game you have got to publish therefore you can't afford to do anything (anything at all) that could risk not being able to, which means safe, guaranteed data generation. Scientific discovery is secondary. Until there's a change from the all consuming obsession with numbers of publications being the single most important thing for a scientists career then things will remain just as they are, if not worsen.

I recently shifted my career from microbiology to systems biology. The thinking end of biosciences in the UK is becoming dominated by computer science. Data analysis, modelling, simulation, and subsequent hypothesis generation are increasingly being given to computer science over biological sciences, who have allowed themselves to drop their numerical / analytical abilities. Linear algebra and quadrattic programming for skills such as flux balance analysis are hugely lucrative in biotech start-ups modelling metabolism. I think ordinary differential equation modelling of biological interactions isn't going anywhere, but statistical modelling for clinical trial design using non-linear mixed effects modelling is enormously lucrative. Optimization for data fitting is also a handy skill set to drop into these as well. Statistics, maths, and computer science graduates going into clinical research organizations can expect to earn 3x the salary of a biosciences graduate going into a lab, and the availability of jobs is significantly higher. Typically they're asking for programming skills in C, Matlab, R, Python, and Java. Bioinformatics roles mining databases is Java, Perl and R and involves database design and graph theory. Modelling and simulation is all C and Matlab, with Python gaining popularity over Matlab due to cost. I've used Mathematica a bit, but Matlab for most. My colleagues all code in Matlab, R or C. Image analysis is also becoming important as high throughput phenotypic screening is in vogue. The people I know in this area are using tools like Matlab and Definians. You will need a PhD in computer science to land the big paying jobs in pharma, and the PhD research will need to be based on biological data of some sort, but the association can be very loose as long as you can code and pick up basic biology along the way. Alternatively, a solid portfolio of projects is also tempting industry due to the lack of skills on the market, and could supplement an M.Sc instead of investing time in a PhD. Personally, I'm seeing a golden age for computer science and maths graduates earning £40-60k straight out of a PhD. Wet lab scientists are starting £16 - 23k, and are increasingly relegated to generating data for computer scientists who are leading the projects. If I had my time again I would train in computer science and see if I could get into the statistical modelling for clinical trials. Do that for a few years, then go freelance and watch the money roll in.