«

»

Jan
20

Thrifty computational science

Can we be thrifty and still do good science? We demand more. For example why validate a model with 20 compounds when you could use 10? It’s all about statistics and power, but really do we over do it. For example from experience grant reviewers want to see you compare other machine learning models, even though you may have found one approach that has been heavily validated and cross validated over several years. So you embark on many hours, weeks or months doing the comparisons to find that really the differences are small.

People tend to be overly cautious why use a cheap computational model when you can do the experiment. This is just a philosophy that seems to permeate our scientific culture. will we ever learn to think and model before we do the experiment. perhaps this could be something the NIH encourages. Thrifty science, is not cutting corners, it’s thinking how to get more bang for the scientific buck.

I was reminded of this when a freshly minted assistant Prof at a Chinese University came to visit last week. We spent a whole day catching up and discussing how she had the same focus to get good peer reviewed papers and yet she was a computational scientist. How was she to achieve this? She had constraints, she has to work in a system, she has no graduate students yet, and very little money. What could she do? I have proposed working on rare diseases because honestly few people are using computational approaches in rare diseases, there are few researchers in many of them anyway, and who knows what new insights she could provide modeling the proteins, docking drugs or other compounds. I am sure whatever insights she could come up with would be greatly appreciated and novel.

And that lead me to think if we can do academic science in a thrifty manner what can pharma learn from this to cut costs. It’s one thing to just pass the work overseas but maybe with slimmed down R&D budgets we could keep jobs in the US by being thrifty. Pharma – does it really need the latest machinery to function? Many start ups exist quite well with used or cast off machinery. Some of my academic friends see it as a coup to snatch up some ex pharma castoff for pennies on the dollar or free. Why can’t pharma just learn to do the same. From personal experience when flat screen computer monitors were just introduced in the 1990s I remember the domino effect as one lab after another spent upto the amount allowed and 1 by 1 replaced their old portable TV style with a new flat screen. The me too effect was real, and the pile of perfectly useable monitors lined up outside labs to be taken away. I am pretty sure this was repeated in department by department, pharma after pharma. This was not Thrifty. I am sure there are other examples?

I have not seen much discussion on thrifty science, possibly something to expand in future.

Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>