Rosetta Abinitio Tutorial

[@mihazeemdeen3713]

It is difficult to find the good tutorials for the Rosetta but your channel useful for rosetta community and protein designer. Thank you very much for your service.

[@BasicScienceSeries]

Sari Sabaan, thank you for this fantastic video. You have made everything clear on the subject! You are a great teacher and an amazing researcher. Please keep posting videos on similar topics.

[@sarfrazgrazian8835]

Can you please make a video on noncanonical amino acids... making params and rotamer files and incorporting them in rosetta

[@gerardhost3821]

Thank you so much for this well explained and articulate tutorial!!! You saved me hours of research! :)

[@TheScientist777]

Can you ignore a conformation that is energetically unfavorable? With the idea that it will eventually reach a more favorable conformation later on in the sequence/folding? Like when aromaticity is broken and restored?

[@anshumansahu1087]

Hello Sari Sabban. Your tutorial has indeed been a blessing. I want to model a transmembrane protein. However, there exists no crystal structure of this protein's extracellular domain (SENSOR DOMAIN) in PDB, making homology modeling extremely difficult. I am pursuing my dissertation and it appears I am at my wit's end. Is it possible to communicate with you through any medium as I believe your thoughtful insights and guidance would be extremely useful. Thanks again for this wonderful tutorial.

[@madhurisatish956]

Thank you for the video! Need help in resolving the issue I get when I Run Rosetta Abinitio on a local computer (Linux) Stage 1 Folding with score0 for max of 20000 protocols.simple_moves.FragmentMover: [ERROR] BEGIN:111 SIZE: 9 TOTAL_RES: 117 protocols.simple_moves.FragmentMover: [ERROR]Are the fragments compatible with the fasta or the input PDB used to extract the folding sequence? protocols.simple_moves.FragmentMover: [ERROR]It appears that the fragments go up to the residue 119 while the pose only has 117 residues! ERROR: Assertion failure: runtime _assert( (begin+size+1)

[@DikshaMarwaha-c2m]

i am trying to predict 3D structure of my small proteins which are 7 to 8000 fasta sequences but they have no homolog template and no pdb. i have to do their ab initio modeling from scratch. how can we do it. do you have any video on it. how can we make fragment library for it

[@visalkavinda]

Fantastic video! Thank you.

[@shahzaibshahid8178]

Hi Sari Sabban, This tutorial is very helpful as I am a beginner. I downloaded Rosetta but couldn't run it on my computer. Kindly help me in this regard. Thank you in advance.

[@sinhasuman1944]

Hi Sari Sabban, immensely good tutorial. Thank you for this. Is it possible for you to make an extensive tutorial for designing a protein/peptide based inhibitor using Rosetta ? That will be great. Thanks again.

[@SariSabban]

Sinha Suman I would love to. I have some other very useful protocols to show in computational biology, I don’t use just Rosetta, I use other algorithms and programs too. I will make a tutorial on them as I find time, but some projects are still not yet fully evaluated because I lack some funding to finish up the experimental parts, such as testing computationally deigned vaccines etc... but hopefully if I ever finish them I will defiantly explain the full process here. Thank you for your comment, I should be publishing a tutorial soon on molecular docking (not using Rosetta) but I will explain my protocol and how I streamed the pipeline, so this is something you might find useful 🙂

[@sinhasuman1944]

@@SariSabban ... sure, will be waiting for that...hopefully, i also will be able to contribute in the discussion in the comments section, I have some experience in small molecule as well as protein protein docking

[@daniellapretorius3259]

Hi, thanks for such an amazing video on both the theory and actual workings of the abinitio protocol! In terms of fragment generation - I have a protein that SS prediction constantly gets incorrect (it is comprised of beta-sheets but the SS prediction considers it coils and loops in Robetta). This incorrect SS prediction gives me bad fragments for my protein. Do you have any recommendations like local fragment production?

[@SariSabban]

Try this bioinf.cs.ucl.ac.uk/psipred/

[@daniellapretorius3259]

@@SariSabban Thank you for this! This does produce a better result but I would have to locally produce my fragments to incorporate this rather than robetta?

[@SariSabban]

@@daniellapretorius3259 implementing fragment generation locally is a bit complicated if you do not have time to implement you can use the server implementation.

[@sri661996]

Hi Dr. Sabban, any plans to make a video on AlphaFold and RoseTTA Fold? Most importantly on the interpretation of the predictions?

[@SariSabban]

Been meaning to do such a tutorial but currently drowning in teaching responsibilities. Hopefully in a month or two when the term ends (and after finishing up several papers) I might have some time.

[@sri661996]

@@SariSabban that would be great, Thank you.

[@joeltorres2158]

Thanks for this tutorial! I have two questions: How could I use Rosetta for a protein with 6 chains? Because the fasta file separated them and How can I create fragments, put on all sequences ignoring the chains or each one chain?

[@SariSabban]

This would really depend on your protein. If the protein is actually one chain with. Issuing segments you might want to model these segments to get a single chain. Or if they separate domains you might want to model each domain separately. It really depends on your particular topic.

[@bluegillsunfish3028]

thank you sari, you really help me a lot!

[@simonbarrera6951]

Hello! If I take a silent file as an input, does the software take the structures within the file for carrying on the simulation? If not, is that any way for doing this? Thank you so much, very nice tutorial

[@SariSabban]

Simón Barrera I’ll be honest with you I haven’t tried using a silent file as input. Try using the flag -in:file:silent ./SILENT_FILENAME I can’t try at the moment since the HPC I am using is being serviced. But if you try please post your result here.

[@jhyang6703]

It's really a very good tutorial for Rosetta Abinitio ,it help me a lot,thank you very much.But i just have a question,in the pbs job script,why not output 25,000 results in one time,but running the script 1,000 time,each time output 25 results?and I can't use #PBS -J options,it returns that "The -J option can only be used in conjunction with -P".

[@SariSabban]

杨嘉辉 We run a 1000 separate jobs each outputting 25 structures for speed. If we run 25 structures (5 minutes per structure) it takes around 2 hours to finish the job. If we run 25,000 structures (5 minutes per structure) it takes around 2,000 hours to finish the job!!! Do you want it wait 2,000 hours (83 days) to get a result? The solution is to setup 1000 separate jobs (each outputting 25 structures) and then “”run them in parallel””, run them all at the same time. Thus we get 25,000 structures in around 2 hours. It’s economics of scale, and can only be done properly in a supercomputer. I hope my explanation was good. :-)

[@jhyang6703]

Thank you very much for answering my question! I understang it,but my supercomputer have only 30 cpus,each cpu have only 16 cores,I think maybe it's hard to get so 25,000 results in a short time.Thans again!

[@jhyang6703]

Thank you very much for answering my question! I understang it,but my supercomputer have only 30 cpus,each cpu have only 16 cores,I think maybe it's hard to get so 25,000 results in a short time.Thans again!

[@teslamooerccx2909]

Thank you so much for the tutorial ! however I have some question , why pyrosetta needs a .pdb file ? I want to get the tertiary structure of a new protein, so I can't give the .pdb file to pyrosetta, what should I do?

[@SariSabban]

I do not understand. You mean you want to predict the structure of a protein?

[@teslamooerccx2909]

​@@SariSabban thanks for your repeat ! yes, I have many small proteins needed to make structural predictions , I only have their sequence, I hope to get a PDB file, can this software implement? how should I do?

[@allisonjones8855]

Thanks a lot, but why can't I complete the clustering? It says it was killed, so I get no result.

[@SariSabban]

Very difficult to answer. I need more information about your steps and setup.

[@afatima121]

Hi Sari Sabban I have a query that my seq is of 856 aa in length and it has only 17% similarity with PDB structures. How can I predict its model using Rosetta? Need your early response.

[@CarlosAndre-nr9em]

If my sequence had 20 aa? How can I do the fragments? Because the Robetta you can input between 27 - 60 aa.

[@SariSabban]

Yes unfortunately the fragments do not work if they are less than 40 amino acids. You can use other structure prediction methods for such small structures, such as Swiss Model. The reason under 40 amino acids does not work is as follows: fragments are the backbone torsion angles (phi and psi angles) of the sequence found in previously solved structures. So what Roberta does is take your sequence and chunks it up into pieces (3 amino acid pieces to get the 3-mer fragments and 9 amino acid pieces for the 9-mer fragments) and searches a database for the torsion angles of the sequences of each piece. The search is done through PSI-BLAST which is very bad at finding homologous between exceptionally short sequences (if you read about PSI-BLAST you will understand). Ab Initio is bad at folding short sequences anyway. So even if you manage to get fragments using other methods the simulation itself will not be very reliable. This is one of its limitations. I hope this explains it. Robetta FAQ answers this too if you want greater details.

[@learner4408]

Do you know online ways to learn complete rosetta course? Thanks

[@SariSabban]

www.rosettacommons.org/docs/latest/Home

[@learner4408]

@@SariSabban thanks sir. Any book or details literature or manual exist as several problems i am facing as a bigger and as such no such experienced person having idea of Rosetta is available in my Institute. I am a PhD student. Thanks again

[@SariSabban]

@@learner4408 I am not aware of a book to be honest, but the demos and the forum are your best bet at the moment.

[@user-sm5ww2ne9x]

@@learner4408 My best recommendation is to reproduce toy examples from the rosetta documentation and slowly adjust them to fit your use case. The videos on this channel are a good start and so are other tutorials, but performing the computational work yourself along with debugging is essential to fully develop your understanding. And also you can always post your issues to the Rosetta forums and devs will answer your questions relatively quickly.

[@jeanpierreramos5320]

Nice video, I was trying to create the fragments locally, but I am finding it difficult, could you help? my peptides are short PS: I'm a beginner

[@SariSabban]

Jean Pierre Ramos I strongly recommend you do not generate fragments locally. The setup is very complicated and requires software that is not available with Rosetta. It is easier if you just submit it to the Robetta server and download the results.

[@jeanpierreramos5320]

@@SariSabban yes, but Robetta online working with peptides >27 aa and my peptides are shorts ( 7 aa)

[@SariSabban]

Jean Pierre Ramos the fragment generation protocol might reject such a small peptide size. Since one is the fragment sizes is 9 amino acids long. But Could be wrong. You might need to ask in the Rosetta forum for clarification rosettacommons.org/forum

[@rajroy2426]

Dr Sabban, Can you give me some keywords to suggest how to do prediction without the structure?

[@SariSabban]

I am not sure what you mean, but in this the protocol requires a protein and a molecule structure in order to simulate the docking using physical principals.

[@rajroy2426]

Dr @@SariSabban At 12:20 , you mentioned that it can be done without the structure.pdb (already folded protein) but it is difficult and you will not be discussing that. I was wondering about that method

[@SariSabban]

I apologise I mistaken your comment for another video. Your best bet is to use the Robetta.org server. It is free. But remember the output isn’t 100% guaranteed to be the real structure of course.

[@rajroy2426]

Dr @@SariSabban thanks

[@高波-e2y]

Hello，sir. I have installed the rosetta，but I follow you to check the Abinitio ，but say Bash no such file , what’s problem, how could I solve?

[@SariSabban]

What operating system are you on?

[@exodus1224]

Hi, I am having this issue as well except with the scoring function. I think it is an issue with our environmental variables but I have no idea how to fix it. Sari please help.

[@SariSabban]

@@exodus1224 which operating system are you working with? Windows, macOS, Linux?

[@kamilkaya5367]

Hello Sir. How can I use Rosetta's energy function for my Proteing Folding Problem using Genetic Algorithm Project? I got stuck on Energy function things.

[@SariSabban]

You can use the energy function as follows: from pyrosetta import * from pyrosetta.toolbox import * init() TheFile = 'THEFILE.pdb' scorefnx = get_fa_scorefxn() pose = pose_from_pdb(TheFile) score = scorefnx(pose) print(score) You will have to modify this script to fit your need

[@ropon-palaciosg.7760]

Hi, rosseta tun without "-s and -navite" structure?,

[@SariSabban]

ROPÓN-PALACIOS G. You can but you will not be able to evaluate the success of the folding. You will basically be folding blindly.

[@ropon-palaciosg.7760]

I'm try modelling small epitopes and designed protein to 50 - 100 aa of legth, this is possible use rosetta?

[@SariSabban]

ROPÓN-PALACIOS G. Yes

[@ropon-palaciosg.7760]

@@SariSabban you can say me how do this based in your tutorial

[@subhrodeepsaha9245]

Hi, thanks for the amazing video! I have just started learning Rosetta and I followed your instructions, it says - ERROR: Unable to open file: /home/rosetta/main/database/chemical/residue_type_sets/fa_standard/residue_types.txt ERROR:: Exit from: src/core/chemical/GlobalResidueTypeSet.cc line: 149 after i entered the command. What could be the solution for this?

[@j.j.gallego3461]

So if I want to fold a protein that is not in PDB...this method is not the best way to do it...isn't it?

[@j.j.gallego3461]

Actually, maybe you have any recommendations. I need to check a lot of proteins from a database and check if they have transmembrane domains and signal peptides... But many are not solved proteins...

[@SariSabban]

It is mainly to simulate the folding of designed proteins, which is why you compare the results to a reference structure (your computationally designed structure). You can still use this protocol for folding structures not yet solved and submitted to the PDB, but that requires little changes to the protocol that I haven’t talked about. You can however try Robetta.org server which is designed for this purpose.

[@j.j.gallego3461]

@@SariSabban Thank you very much! I am trying Robetta, It could be really nice to find a tutorial, robetta server is oc way more simple than all the command line work, but still, there are a few options long waiting times if you do something, bud, I will keep looking for tutorial and trying. Thanks a lot for your quick reply :)

[@ariellee9428]

@@SariSabban Hey thank you for the video. I wonder what's the difference between running Rosetta abinitio vs using robetta online server?

[@khushwantsingh6870]

what if we do not have PDB structure of a protein

[@SariSabban]

khushwant singh that is slightly more difficult. You can still attempt to simulate its folding, but you will have to find similar structures to it, it’s called “tethering” in order to generate the fragment files required for Abinition. You can try your luck we Roberta.com you can submit a FASTA sequence for structure determination, it will take a while, and it will only simulate one domain at a time, then it will print you 5 structures. See if they are similar. But without a reference PDB structure always take these simulations with a grain of salt.

[@vasavigarisetti3064]

Sari Sabban If my sequence has disordered regions and likewise no structure because of that property would ab initio structure prediction be helpful?

[@abhignanu8922]

hii, how can i use other tools of rosetta?

[@SariSabban]

Abhigna N U N U each tool is widely different in its parameters and evaluation. You will have to look for the demos of each tool you want to use.

[@smbat6828]

A good place to start would be with the Meiler Lab tutorials. Link given below. www.meilerlab.org/index.php/rosetta-tutorials

[@rogerj9425]

Hi Sari, great videos. Can this be done using Pyrosetta in Google Colab?

[@SariSabban]

Roger J I haven’t tried it. I know PyRosetta does not work with a GPU. If you get it working in colab please let me know here.

[@SariSabban]

RJ are you able to execute the binary? The abinitio binary? If you can then try running a simple (-nstruct 10) using the CPU and again using the GPU and see if there is a difference. How to ensure that PyRosetta is using the GPU I don’t know, you will have to search. But please post your results here. A lot of people might find this useful.