Wednesday, March 27, 2013

Hypothesis Generation in Biological Research Grows

Titus Brown wrote a long, yet wonderful, narrative of how the ease of doing 'discovery science' and generating tons of data (by sequencing or other 'omics approaches) is rapidly leading to extended periods of analysis prior to doing more experiments:
Hypotheses are now cheap (because data is plentiful) and I think we should focus on developing likely hypotheses and winnowing out the dumb 'uns computationally before we ever pick up a pipette man to test 'em. That is, expand the hypothesis-generation and analysis stages so that we're more likely to develop a comprehensive and interesting hypothesis.
Most computational biologists have experienced the tension between needing to analyze data and actually doing more experiments to test ideas.  The underlying source of that tension is usually another scientist that believes in a mistaken assumption: Doing experiments counts as work; Data analysis does not.

Thankfully, it's also an idea that's rapidly becoming anachronistic. 

More and more biologists (i.e. bench scientists) are becoming dependent on people programming and using computers in ways that a regular desktop users does not.  Using Excel to do bioinformatics doesn't cut it.