Ask HN: What biology research projects should I pick?

by christopher8827 on 6/24/2023, 2:15:08 PM with 1 comments
I have been offered a list of biology + data science research projects at my university. I do aim to get into Deepmind or some other AI / biotech company in the future or at least try to commercialise some of my research. What do you think have the highest potential to contribute to society and advance science & the goals above?

- Increasing the predictive accuracy of an ML model that predicts whether using a infrared spectral signature to identify if an individual has a mosquito-related disease like malaria. eg. would be used in a scanning device to replace PCR.

- Prediction of bacterial species resistance to the antibiotic rifampicin. This would use ML to build a model with genome data and predict antibiotic resistance in certain bacterial species and whether some species are unable to evolve resistance. This data would be used in future, experimental projects.

- Improving scalability of single-cell immune repertoire analysis. Omniscope is one startup working on improving current workflows.

- Developing better statistical methods to assess diversity of immune receptors using sequencing depths or cell numbers

- Single-cell trajectory analysis with immune repertoires

  • by COGlory on 6/24/2023, 2:44:58 PM

    Biology is a tough nut to crack with boatloads of computational irreducibility, some of which we know about, some of which we don't.

    I've seen so many problems get hung up on two things:

    1) It was a SNR problem at the point of data measurement, and no amount of ML can fix that.

    2) It's a computational irreducibility problem that was mistaken for a convolution

    Of what you listed, the penultimate honestly sounds the most realistic and least pie-in-the-sky. Personally, I'd avoid anything predictive because of the two issues I mentioned. It can be done for some things, but it's really quite difficult to tell for what.