AlphaFold2 Full Tutorial | Introduction to ColabFold For Wet Lab Researchers

Let's take a first look at AlphaFold2 and how to practically use it as a wet lab biologist in under 100 seconds. AlphaFold2 is a highly accurate protein structure prediction model written by Google's DeepMind and in this tutorial we show you just how to use it using the Neurosnap platform.
#biology #biochemistry #proteinstructure #molecularbiology #deeplearning
Check out Neurosnap: neurosnap.ai/
AlphaFold2 Blog Posts:
Part 1 - neurosnap.ai/blog/post/641a34...
Part 2 - neurosnap.ai/blog/post/642224...
Part 3 - neurosnap.ai/blog/post/642243...
See the results from this job at neurosnap.ai/job/64fe68141bbe...
See the protein used in this experiment at www.uniprot.org/uniprotkb/I1E...

[@baka6884]

How can i see the progress like in alfafold?

[@zainiiBee]

which is better, esm fold or alpha fold?

[@NeurosnapInc]

Both are good, AlphaFold2 is almost always better, ESMFold is supposedly a bit more accurate on smaller monomers. Personally, I suggest using both as they're both very fast and you can compare the results afterwards to see which prediction is more rationale.

[@vpanagiotou481]

i have one question sir.
I ran alphafold2 in google colab for a sequence of 988 amino acids, and it took about 1,5 hour to finish.
Here in neurosnap platform i ran the same amino acid sequence, and it took about 30 minutes to finish.
Did neurosnap platform apply any optimizations in the source code of alphafold2 in order to be faster ?
Or does it have GPUs that are more powerful than google colab's? (in colab i used an A100 GPU with the free-user option, i did not subscribe to the Colab Subscription Pricing.)
I study the algorithmic part of Allphafold2 and ESMFold, so it is very important to know if there is any such code optimization in neurosnap algorithm implementations.
Is it the same to execute this algorithms either in colab or neurosnap, or should i take into consideration the fact that there have been some code optimizations and changes? Is the source code of colab and neurosnap the same?
are the neurosnap algorithms based on github code, like colab?

[@NeurosnapInc]

We use the ColabFold implementation in our backend which is substantially faster than the AlphaFold2 implementation provided by Deepmind. Also, make sure to check your input parameters like ensembles and recycles as well as both will have a noticeable impact on prediction time.

[@hkl6798]

So no need to use colab?

[@NeurosnapInc]

That's correct. There is no need to use colab as everything can be done directly on the neurosnap platform.