4 ms·
People already can grow up billions of cells for experiments and use supercomputers to process terabytes of sequencing data with statistical modelling. The hard
by asdff 5y ago
People already can grow up billions of cells for experiments and use supercomputers to process terabytes of sequencing data with statistical modelling. The hard part isn't the scale, but designing experiments and figuring out what evidence is needed to answer specific biological questions. That stuff you can't speed up with robotic arms and more processor cores, it takes time for people to think about these things and have conversations with others about these topics.
- EamonnMR 5y agoNot to mention the amount of time experiments take-suppose we want to know if a drug prevents an illness that takes decades to develop? Sure you can model it in a mouse, but to do that you need to know exactly what the parameters of your model are.
- amelius 5y agoI'm not so sure about that. For example, with robots you can run "for GENE in GENOME do ..." For humans even thinking about what that statement does is a lot of work already!
- mythrwy 5y agoIt's not a one dimensional problem though. There is a lot of interaction between the various systems in a human body. To keep with the programming analogy, ya, you can see the SUM statement and that is necessary to understand, but what is it for? What is it's role in the larger program? Where do it's inputs come from? Where do they go? What effect does changing it have in other places? I do agree our knowledge of genetics etc. will be much expanded soon. But we'll probably also make some nasty errors.
- amelius 5y ago> There is a lot of interaction between the various systems in a human body. Yes, the interactions are the interesting bit. Perhaps we can figure out the biological pathways at the cellular level by taking a systematic approach of turning genes on/off and turning their expression on/off, then looking at the expression of other genes. This is exactly what could be done by robots.
- ethanbond 5y agoI think you’re seriously underestimating the scale of the genome and the number of possible combinations that might be responsible for some phenotype, even assuming the genotype -> phenotype causality was straightforward (which it’s not). This is simply not a problem we’re going to computer our way through, at least until we have large scale quantum computers and way better data than we have today.
- jcims 5y agoNaively, a small protein of ~100 amino acid residues has ~21^100 configurations. Amino acids average 110 Daltons each (110g/mol), you would need approximately 3*10^107kg of matter to build them all. There is an estimated 10^53kg of matter in the known universe.