<html><head><meta http-equiv="Content-Type" content="text/html; charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">Hi Raihan,<div class=""><br class=""></div><div class=""> The maximum sequence length that can be predicted by AlphaFold is limited by GPU memory. The following table says that paid Google Colab services only offer at most 16 Gbytes of GPU memory (from <a href="https://blog.paperspace.com/alternative-to-google-colab-pro/" class="">https://blog.paperspace.com/alternative-to-google-colab-pro/</a>). Not sure if it is correct since the table is from a Google competitor.</div><div class=""><br class=""></div><div class=""><img apple-inline="yes" id="3F14B4D8-5270-4BC1-AFDE-6B36DE38025E" width="544" height="480" src="cid:9E3B2B15-2AF6-48D5-ADE5-29E8E9F91AC0" class=""></div><div class=""> You can run sequence length 1000, sometimes 1200, with 16 GB of GPU memory as shown here</div><div class=""><br class=""></div><div class=""><span class="Apple-tab-span" style="white-space:pre"> </span><a href="https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html" class="">https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html</a><br class=""><div><br class=""></div><div> Various cloud compute services offer modern GPUs like Nvidia A40, A100, A6000 with 48 GB, or 40 or 80 GB of GPU memory usually charging about $2-3 per hour. But ChimeraX can only currently use Google Colab for AlphaFold predictions.</div><div><br class=""></div><div><span class="Apple-tab-span" style="white-space:pre"> </span>Tom</div><div><br class=""></div><div><br class=""><blockquote type="cite" class=""><div class="">On Oct 31, 2022, at 2:45 PM, S.M. RAIHAN RAHMAN <smrrbtgeiu@gmail.com> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class="">Thank you. I saw the run times. Is it possible to run sequences of around 1200 in Google Colab pro plus paid service through ChimeraX? In the website link you shared, I only saw about Google Colab pro paid service. Please let me know. <div class=""><br class=""></div><div class="">Regards,</div><div class="">Raihan</div></div><br class=""><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, 31 Oct 2022 at 14:33, Tom Goddard <<a href="mailto:goddard@sonic.net" class="">goddard@sonic.net</a>> wrote:<br class=""></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div style="overflow-wrap: break-word;" class="">ChimeraX only runs AlphaFold on Google Colab. Because Google Colab provides 5 year old GPUs with little memory (16 GB) it can only run sequences of length about 1000 or less. With better equipment and your own AlphaFold installation you can run sequences up to length 3000 or 4000 although that can take 30 hours. Here are example run times<div class=""><br class=""></div><div class=""><span style="white-space:pre-wrap" class=""> </span><a href="https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html" target="_blank" class="">https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html</a><br class=""><div class=""><br class=""></div><div class=""> Tom</div><div class=""><br class=""><div class=""><br class=""><blockquote type="cite" class=""><div class="">On Oct 31, 2022, at 2:28 PM, S.M. RAIHAN RAHMAN <<a href="mailto:smrrbtgeiu@gmail.com" target="_blank" class="">smrrbtgeiu@gmail.com</a>> wrote:</div><br class=""><div class=""><div dir="ltr" class="">Thank you so much. Is there any alternative to Google Colab for running AlphaFold through ChimeraX? Please let me know.<div class=""><br class=""></div><div class="">Regards,</div><div class="">Raihan</div></div><br class=""><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, 31 Oct 2022 at 14:22, Tom Goddard <<a href="mailto:goddard@sonic.net" target="_blank" class="">goddard@sonic.net</a>> wrote:<br class=""></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi Raihan,<br class="">
<br class="">
The ChimeraX AlphaFold prediction produces 5 models but it only automatically opens the best model. All 5 models are in the ~/Downloads/ChimeraX/AlphaFold/prediction directory after the job completes. There is no option to run fewer models. Running AlphaFold not through ChimeraX also produces 5 models. This is because AlphaFold works by running 5 slightly differently trained neural networks and the best is chosen. If you install your own AlphaFold (a somewhat difficult task, needs Linux and an Nvidia GPU and a few Tbytes of disk for databases) then you can modify the AlphaFold Python code to predict only one or a few models by choosing which of the 5 networks it should run.<br class="">
<br class="">
Tom<br class="">
<br class="">
> On Oct 31, 2022, at 1:39 PM, S.M. RAIHAN RAHMAN via ChimeraX-users <<a href="mailto:chimerax-users@cgl.ucsf.edu" target="_blank" class="">chimerax-users@cgl.ucsf.edu</a>> wrote:<br class="">
> <br class="">
> Hi!<br class="">
> Good Afternoon. I hope that you are doing well. I am facing a problem in running structure prediction of large protein sequence in AlphaFold prediction. Is there any way of predicting only 2 models of a given sequence in lieu of 5 models before providing the final model in AlphaFold prediction? Please let me know. Thank you. <br class="">
> <br class="">
> Regards, <br class="">
> Raihan<br class="">
> _______________________________________________<br class="">
> ChimeraX-users mailing list<br class="">
> <a href="mailto:ChimeraX-users@cgl.ucsf.edu" target="_blank" class="">ChimeraX-users@cgl.ucsf.edu</a><br class="">
> Manage subscription:<br class="">
> <a href="https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users" rel="noreferrer" target="_blank" class="">https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users</a><br class="">
<br class="">
</blockquote></div>
</div></blockquote></div><br class=""></div></div></div></blockquote></div>
</div></blockquote></div><br class=""></div></body></html>