Kansas Crimson

TraditionKU wrote: ↑Fri Jul 16, 2021 4:19 pm

PhDhawk wrote: ↑Fri Jul 16, 2021 3:54 pm

TraditionKU wrote: ↑Fri Jul 16, 2021 3:41 pm
speed and specificity?

i’d honestly gave to look more closely at them all, but i’d assume that the two new AI technologies have outperformed those that you cited in the relevant competitions (presuming they participated in those or similar competitions)

are those other three open source as well?

from what i’ve read, at least with AlphaFold, it does a much better job of replicating the results of x-ray crystallography and cryo-electron microscopy, but does so with a fraction of the time and cost

They're open sourced, I'm not sure that they use AI.

I just don't know how you can trust a computer derived 3d structure de novo.

I mean, I think it's great for, let's say the spike protein on the coronavirus, which has been crystalyzed gets another mutation and you want to see how that changes the protein structure...that's great, you have a REAL structure to start with, the computer gives you the most likely structure based on changing 1 or a few amino acids. Fast, cheap, reliable. On the other hand, lets say there's a new pathogen that shows up tomorrow and they want to determine the structure of a completely novel protein...sorry, but I'm only going to put so much trust into a structural prediction, until it gets its structure determined experimentally.

completely agree

but i think this tech at least gives a starting point in such a case

in what i do, where epitope variation can drastically affect the efficacy of an antibody therapy, it’d be good to know where those differences are likely to be without investing the time and money into experimental procedures unnecessarily...particularly in a predictive way

Yeah, totally.

Just was curious why this specifically warranted a Nature article when there are lots of 3D structure predictors. I guess it must be a big advance in quality or the technology used.

Wasn’t me.

here’s something positive for this thread

pretty amazing, and yet another reason why biodiversity matters

https://advances.sciencemag.org/content/7/29/eabf9635

PhDhawk wrote: ↑Fri Jul 16, 2021 5:23 pm

TraditionKU wrote: ↑Fri Jul 16, 2021 4:19 pm

PhDhawk wrote: ↑Fri Jul 16, 2021 3:54 pm They're open sourced, I'm not sure that they use AI.

I just don't know how you can trust a computer derived 3d structure de novo.

I mean, I think it's great for, let's say the spike protein on the coronavirus, which has been crystalyzed gets another mutation and you want to see how that changes the protein structure...that's great, you have a REAL structure to start with, the computer gives you the most likely structure based on changing 1 or a few amino acids. Fast, cheap, reliable. On the other hand, lets say there's a new pathogen that shows up tomorrow and they want to determine the structure of a completely novel protein...sorry, but I'm only going to put so much trust into a structural prediction, until it gets its structure determined experimentally.

completely agree

but i think this tech at least gives a starting point in such a case

in what i do, where epitope variation can drastically affect the efficacy of an antibody therapy, it’d be good to know where those differences are likely to be without investing the time and money into experimental procedures unnecessarily...particularly in a predictive way

Yeah, totally.

Just was curious why this specifically warranted a Nature article when there are lots of 3D structure predictors. I guess it must be a big advance in quality or the technology used.

an update:

https://deepmind.com/blog/article/putti ... rlds-hands

TraditionKU wrote: ↑Tue Jul 27, 2021 1:51 pm https://techcrunch.com/2021/07/22/deepm ... alphafold/

I remember when next gen sequencing was developed when I was in grad school and they bought both a 454 and an illumina sequencer, and then within a couple months, the people who worked in the sequencing facility who did all the old Sanger sequencing were shortly laid-off cause that technology was so cheap that it was cheaper to send off for sequencing than to do it in house. Felt bad for the people who lost their jobs, but I guess that's what happens when your technology becomes outdated.

I wonder if there are a lot of protein crystallographers looking for new lines of work/research.

PhDhawk wrote: ↑Tue Jul 27, 2021 1:54 pm

TraditionKU wrote: ↑Tue Jul 27, 2021 1:51 pm https://techcrunch.com/2021/07/22/deepm ... alphafold/

I remember when next gen sequencing was developed when I was in grad school and they bought both a 454 and an illumina sequencer, and then within a couple months, the people who worked in the sequencing facility who did all the old Sanger sequencing were shortly laid-off cause that technology was so cheap that it was cheaper to send off for sequencing than to do it in house. Felt bad for the people who lost their jobs, but I guess that's what happens when your technology becomes outdated.

I wonder if there are a lot of protein crystallographers looking for new lines of work/research.

i experienced the same

my masters work was with Sanger-derived, multi-locus datasets, but phd had a combination of Sanger and nex-gen datasets

i’d think, tho, that there are going to still be x-ray crystallographers out there either proofing the AI estimates, or working on the proteins that the AI had low confidence in. but yes...this is gonna push plenty of people into new roles

TraditionKU wrote: ↑Tue Jul 27, 2021 1:58 pm

PhDhawk wrote: ↑Tue Jul 27, 2021 1:54 pm

TraditionKU wrote: ↑Tue Jul 27, 2021 1:51 pm https://techcrunch.com/2021/07/22/deepm ... alphafold/

I remember when next gen sequencing was developed when I was in grad school and they bought both a 454 and an illumina sequencer, and then within a couple months, the people who worked in the sequencing facility who did all the old Sanger sequencing were shortly laid-off cause that technology was so cheap that it was cheaper to send off for sequencing than to do it in house. Felt bad for the people who lost their jobs, but I guess that's what happens when your technology becomes outdated.

I wonder if there are a lot of protein crystallographers looking for new lines of work/research.

i experienced the same

my masters work was with Sanger-derived, multi-locus datasets, but phd had a combination of Sanger and nex-gen datasets

i’d think, tho, that there are going to still be x-ray crystallographers out there either proofing the AI estimates, or working on the proteins that the AI had low confidence in. but yes...this is gonna push plenty of people into new roles

Yeah, there will definitely still be a need for crystallographers, just maybe not as many.

Protein crystallographers will never be out of a job! Pharma industry is full of them and we continue to hire them. They are a valuable resource in drug discovery.

in a half dozen years, this might go in the Great Outdoors thread...

Firm raises $15m to bring back woolly mammoth from extinction

I saw that headline too.

Apparently they think it’ll help combat climate change?

They posit mammoth grazing habits will work to move snow from the tundra ground, which will allow ground exposure to the colder air, which will help to keep the permafrost permanently frozen

wishful thinking, imo. i’ll never understand the frankenstein-like thinking of bringing a species back from extinction, rather than using the resources and know-how to save the ones still living

they also justify their effort by claiming it will help “save” asian elephants. lol, no...their just replacing them with a bio-engineered mutant

https://m.dw.com/en/africa-swine-fever- ... a-59212422

Africa Swine Fever: Is China downplaying another disease outbreak?

... Chinese officials first identified ASF in 2018 in Liaoning, a coastal province in China's northeast. Unlike SARS-CoV-2, ASF does not threaten human health. The disease, however, is highly virulent in pigs and mortality rates approach 100%.

ASF is of particular concern in China, the world's top pork consumer and home to roughly half of the world's pigs. Given the country's enormous pig population, ASF could turn China into a significant reservoir of disease and pose a threat to China's neighbors in the region if it becomes endemic...

i know i’m a broken record on this topic, but humanity really needs a paradigm shift in the way we interact with and utilize other animals

this sounds like the pigs problem. not mine.

i wouldn’t expect you to understand

https://nautil.us/issue/107/the-edge/th ... erapeutics

seems like a waste of time and resources since we’re all gonna die anyway

why go to the effort of improving quality of life?

MICHHAWK wrote: ↑Fri Sep 17, 2021 10:00 am this sounds like the pigs problem. not mine.

JFC dude have a little awareness…don’t you realize what this could do to the bbq thread?!

TraditionKU wrote: ↑Wed Oct 06, 2021 4:25 pm https://nautil.us/issue/107/the-edge/th ... erapeutics

seems like a waste of time and resources since we’re all gonna die anyway

why go to the effort of improving quality of life?

As a small-molecule person I’m amazed by the specificity of these RNA therapies. The biggest issue facing them is delivery - as in how to get it to the right place without degrading it in the process

zsn wrote: ↑Thu Oct 07, 2021 10:03 am

TraditionKU wrote: ↑Wed Oct 06, 2021 4:25 pm https://nautil.us/issue/107/the-edge/th ... erapeutics

seems like a waste of time and resources since we’re all gonna die anyway

why go to the effort of improving quality of life?

As a small-molecule person I’m amazed by the specificity of these RNA therapies. The biggest issue facing them is delivery - as in how to get it to the right place without degrading it in the process

in my field there’s a bunch of excitement around repurposing small-molecule drugs that already have FDA approval for other indications with the hope of reducing the cost-to-market and bypassing some of the clinical trial hurdles. apparently there is already a clinical trial underway for a phospholipase inhibitor

i just saw an update from some colleagues in Denmark that the alphafold2 AI technology is having success in predicting multi-chain protein complex shapes. i’d imagine they’ll leverage the results from their work with alphafold to improve upon their work with phage display for identifying and selecting recombinant human antibodies

so much is happening on the academic side of things that it’s hard to know which platforms/technologies will end up making it to market