Hacker News Reader

Comment by teekert | Hacker News Reader

As a biologist and open source/Linux nerd, I can tell you it is because the biologists don't understand you or what you can do. And they find it annoying that you pretend to understand their field.

Biologists do things "knowledge based", they love to quote stories in which data-driven research ended in the conclusion: "ATP is an important molecule". They love terms like datarrhea. They love the Google Flu fiasco. They know what over-fitting is but never heard of random forests.

And I can tell you, quite often they are right. Computer scientists often don't understand and appreciate the complexities of a signalling pathway, the insane uncertainties, the lack of reality in any of the models biologists use to illustrate connections, the ease with which models are discarded.

So tell me, what are your hypotheses and how do you think your algorithms can prove them right or wrong?

View on HN

>> Computer scientists often don't understand and appreciate the complexities of a signalling pathway

Do they have to? If I understood OP right, he just wants to crunch some numbers and find out whether there is a correlation between specific bacteria types and diseases. It's a stochastic approach to the problem.

I don't like rants like yours that basically say "don't talk about this subject, you don't have the necessary background". Problems sometimes _are_ solved by people foreign to a field.

"he just wants to crunch some numbers and find out whether there is a correlation between specific bacteria types and diseases."

I don't think he is, he is specifically looking for biologists.

I don't like your: "he just wants to crunch some numbers". So what if you find a correlation? You can always find a correlation when the number of parameters is much greater than the number of experiments, that is the problem with knowledge-less Big Data. Suppose you do find a correlation. Is it because of the bacteria? Or is it because of the Genomic similarities between the persons? Is it because those persons were breast fed? Do the hate milk? Or when there is no correlation, is it because the microbiota has no effect? Or is it because the bacteria in one gut are continuously exchanging plasmids? Again, what is your hypothesis?

I like the computer scientists that reach out to other field and help in generating hypotheses. Just crunching some numbers sound like a recipe for nothing.

I mean, when you do find a correlation? Are you going to advice people medically based on a correlation? I know the FDA for one does not like that.

Are you accusing computer scientists of not being able to apply scientific methods? I'm not sure, and I don't want to setup a strawman. Of course there would have to be hypotheses, and they would be tested. Scientific methods are not exclusive to biology. In the study or citizen study I imagine here, a questionnaire would include questions like yours.

Crunching the numbers is the means to refute them, then. I don't know what you think CS folks are doing all day?

Thanks for the enlightening reply! It definitely helps to understand what is going on in the mind of a biologist when they encounter a computer scientist.

I think what draws me most to the field of biology, and the prospect of biological computing, is that there are insane uncertainties and super complex pathways; the fact that the same test run 10 times in a biological machine may not yield the same result each time.

My main challenge is in starting a dialog with people in the field. Perhaps I will try communicating in scientific method lingo and see how that fares.

Overall, the feeling I get is that biologists fear the commoditization of their field, and I observe that even the progressive bio startups are still sticking to the (closed) pharmaceutical model of business.

Reply on news.ycombinator.com