5 ms·
Having studied biology and genetics this is common sense to me. Genetic systems are probabilistic chemical systems where everything influences everything to var
by api 8y ago
Having studied biology and genetics this is common sense to me. Genetic systems are probabilistic chemical systems where everything influences everything to varying degrees in parallel. They are nothing like the linear assemblies of modules that humans prefer to engineer.
- stochastic_monk 8y agoWhich explains why gene regulatory network inference by expression patterns is garbage. My first time in graduate school I worked on it, and the model worked just as well using random genes instead of annotated transcription factors. There was no science in it, so I took a master’s, worked in industry, and am now doing real science in a CS PhD program.
- fizx 8y agoCommon sense from the machine learning side too! (that said, the ex-physicist in me would like to remind everyone that science and reality are not required to match your common sense) If true, this means that evolution could look more like stochastic gradient descent, and less like brute force luck plus survival of the fittest.
- onychomys 8y agoFrom an evolution point of view, it actually just means that selection happens at the organismal level, not the gene level.
- fizx 8y agoYou misunderstand my point... larger ratio of base pairs to behavior means that the behavior is likely more robust to noise, and the original evolution of the behavior would have been smoother (i.e. proceeded in less of a step-wise manner, because each step would be a much smaller fraction of the required change than previously thought).
- politician 8y agoTraits embedded into a dense vector space encoded as DNA base pairs is a pretty radical deviation from the one-hot vector approach we were taught in school. Wouldn't it be lovely if we found that vector addition worked similarly in this domain?
- aphextron 8y ago>Having studied biology and genetics this is common sense to me. Genetic systems are probabilistic chemical systems where everything influences everything to varying degrees in parallel. They are nothing like the linear assemblies of modules that humans prefer to engineer. I'm really glad that this mental model I've built up seems to reflect reality. The idea of neatly editing specific genes to produce a desired result always sat with me as the pinnacle of human hubris. The more I learn about Biology the more I just throw up my hands and say "magic".
- bookofjoe 8y ago>The idea of neatly editing specific genes to produce a desired result always sat with me as the pinnacle of human hubris. See, for example: Gene Therapy in a Patient with Sickle Cell Disease https://www.nejm.org/doi/full/10.1056/NEJMoa1609677 https://www.nejm.org/doi/full/10.1056/NEJMoa1609677
- aphextron 8y agoI literally can't understand a single word on that page. Care to explain?
- bookofjoe 8y ago"We describe our first patient treated with lentiviral vector–mediated addition of an antisickling β-globin gene into autologous hematopoietic stem cells."
- politician 8y ago"We describe our first patient treated with <a technique that injected> <DNA to counteract the disease> into <stem cells>. <which were then injected into the patient>."
- bookofjoe 8y agoOr you could simply say "We performed a chemistry...":-)
- deegles 8y agoI wonder if years from now we'll look back at the "selfish gene" the way we look at Freudian ideas about psychology.
- api 8y agoDefinitely. I am generally unimpressed by Dawkins.