patrickbryant1 / evobind Goto Github PK

Hi:

I'm trying to use EvoBind from Google Colab but I'm having a problem. When I run the code cell "Run EvoBind" it is supposed to take 7 hours maximum, but I have done 23 hour tests and it has not given me any results. So far the only thing I have obtained is that message from libraries in which you can see in the image which I have attached.

Finally, I tried the tutorial associated with this GitHub and I have the same problem

Hello,
I am trying to use the local version of the software, and what I noticed is that, with both the test case and my interested protein, when I specify a receptor target residue, the peptide location starts on a totally different location (so far so good) but it doesn't chage too much through the iterations.
So for example, on the test case 1ssc.pdb, I have selected the residue 28 as target interface, however at each iteration the peptide is close to residue 108, that is on the opposite side of the protein.

I have also tried asking for a 3 aa peptide, and after 80 iterations the peptide, although having a mutated sequence , it's still there. It's seems to me that the distance between the peptide COM and the receptor target COM it's not taken into account too much.

Any help?

In setup.sh the following command causes an error
singularity pull AF_environment.sif docker://catgumag/alphafold:latest

The error is
FATAL: While making image from oci registry: failed to get checksum for docker://catgumag/alphafold:latest: Error reading manifest latest in docker.io/catgumag/alphafold: manifest unknown: OCI index found, but accept header does not support OCI indexes

the docker pull command works fine

docker pull catgumag/alphafold:latest
latest: Pulling from catgumag/alphafold
eaead16dc43b: Pull complete 
f66018fd6918: Pull complete 
121ce45b8595: Pull complete 
5a5bda8264e7: Pull complete 
b05cc2528b1a: Pull complete 
7a8fd3721580: Pull complete 
7cdd51c189d4: Pull complete 
f43eb09af0d1: Pull complete 
6e59b32a7f85: Pull complete 
0b7e2150641d: Pull complete 
e82ea881eb2f: Pull complete 
4f4fb700ef54: Pull complete 
9c5bc5c2c45c: Pull complete 
Digest: sha256:8e9aaf56e4bbd23cef7d607ec5fee4dfddfa94eab79ff12b43bb31b90eba397f
Status: Downloaded newer image for catgumag/alphafold:latest
docker.io/catgumag/alphafold:latest

EDIT:
this might be a docker hub problem. The following command works
singularity pull AF_environment.sif docker://catgumag/alphafold:2.3.1
but the following commands do not
singularity pull AF_environment.sif docker://catgumag/alphafold:2.3.2 and
singularity pull AF_environment.sif docker://catgumag/alphafold:latest

Hi Patrick,

I have a couple of additional questions.

1. How is the change of amino acid at each loop is defined? My understanding is that it is completely random. However if this is the case the machine has no memory of previous changes. Therefore peptides might be computed twice.
2. How is chosen the amino acid in the peptide sequence that should be modified?

Do you think that a computation of the pIddt of each aa or a running average of it, might be useful to define which aminoacid modify first to improve the binding?

Thank you for the answers, I found the software very nice! It is fun to test it.

Ema

Hi!
I tried to run the sample colab notebook but got this error:

AttributeError: module 'jax' has no attribute 'tree_multimap'

which is apparently a deprecated function. Is there a fix for this?

Thanks!

FATAL: could not open image /dssg/home/acct-clswg/clswg/evobind/EvoBind/python3: failed to retrieve path for /dssg/home/acct-clswg/clswg/evobind/EvoBind/python3: lstat /dssg/home/acct-clswg/clswg/evobind/EvoBind/python3: no such file or director

Hello! I got the following syntax;

NameError Traceback (most recent call last)
in
19 from af_mod_pred import run_preds
20 TARGET_FASTA=OUTDIR+PDBID+'_'+TARGET_CHAIN+'.fasta'
---> 21 if OPTION=='1':
22 TARGET_MSA=OUTDIR+'/str_aln/'+PDBID+'_str.a3m'
23 else:

NameError: name 'OPTION' is not defined

When attempting to use the google colab sheet under the

#@markdown #Evaluate the designed sequences using our modified version of AlphaFold adapted to binder evaluation.
#@markdown The designed sequences are predicted in complex with the target.
#@markdown The predictions are used below to assess the designs by calculating a custom bind-score.

I tried with the example protein as well as my own protein of interest and got the same error.

Any suggestions?

I think the work you have done is super exciting and I hope I can make this work (I intend to validate binders with wet-lab experiments asap).

Best,

Troels

hi, i want to install the Evobind2 by using the environment.yml, but i found that this file seems to be an old version of the file, when i use this file to creat a conda environment, I encountered such a mistake:

LibMambaUnsatisfiableError: Encountered problems while solving:

• package openmmforcefields-0.11.2-pyhd8ed1ab_1 requires ambertools >=20.0,<23, but none of the providers can be installed

Could not solve for environment specs
The following packages are incompatible
├─ ambertools ==23.3 py312h1577c9a_6 is requested and can be installed;
└─ openmmforcefields ==0.11.2 pyhd8ed1ab_1 is not installable because it requires
└─ ambertools >=20.0,<23 , which conflicts with any installable versions previously reported.

thank you!

Hi Patrick,

I have a question more than an issue. I was wondering how do you define the center of mass of the peptide.
In my mind it should be the expected position of the CA of the middle amino acid of the peptide.
However this position is a bit unpredictable especially if you do not start with a know surface.

I was wondering how you guess this parameter and if I an estimating this correctly.

How much this position affects the prediction? In other words, does this change during the evolution?

Thank you for the answer!

Ema

hi i wonder where the parameter of output numbers lie
and i changed NITER to 50 to test and it resulted in 49 unrelaxed structures
is that expexted?

Problem

When I use a structure other than the predefined one, I get this error

NameError Traceback (most recent call last)
in ()
61 #Check if CALCULATE_BINDER_COM
62 if CALCULATE_BINDER_COM==True:
---> 63 TARGET_CAs = receptor_CAs[np.array(TARGET_RESIDUES)-1]
64 BINDER_COM = np.sum(TARGET_CAs,axis=0)/(TARGET_CAs.shape[0])
65 RECEPTOR_CAs = prepare_input(OUTDIR+PDBID+".pdb", RECEPTOR_CHAIN, TARGET_RESIDUES, BINDER_COM, OUTDIR)

NameError: name 'receptor_CAs' is not defined

PDBID = "1q94" #@param {type:"string"}
RECEPTOR_CHAIN = "A" #@param {type:"string"}
RECEPTOR_SEQUENCE = "GSHSMRYFYTSVSRPGRGEPRFIAVGYVDDTQFVRFDSDAASQRMEPRAPWIEQEGPEYWDQETRNVKAQSQTDRVDLGTLRGYYNQSEDGSHTIQIMYGCDVGPDGRFLRGYRQDAYDGKDYIALNEDLRSWTAADMAAQITKRKWEAAHAAEQQRAYLEGRCVEWLRRYLENGKETLQRTDPPKTHMTHHPISDHEATLRCWALGFYPAEITLTWQRDGEDQTQDTELVETRPAGDGTFQKWAAVVVPSGEEQRYTCHVQHEGLPKPLTLRWE" #@param {type:"string"}
TARGET_RESIDUES = "5,7,9,24,25,33,34,45,59,62,63,64,65,66,67,68,69,70,72,73,74,75,76,77,78,80,81,84,95,97,99,114,116,123,124,133,139,140,142,143,144,146,147,152,155,156,157,158,159,160,163,164,167,168,171" #@param {type:"string"}
TARGET_RESIDUES = [int(x) for x in TARGET_RESIDUES.split(',')]
CALCULATE_BINDER_COM = True #@param {type:"boolean"}
BINDER_LENGTH = 9#@param {type:"integer"}
NITER = 300#@param {type:"integer"}
BINDER_COM = "33.966637,23.854908,9.859454" #@param {type:"string"}
BINDER_COM = [float(x) for x in BINDER_COM.split(',')]
RECEPTOR_MSA = "1q94_receptor.a3m" #@param {type:"string"}

I used an a3m from a previous AlphaFold run
How to fix

Change the variable name from receptor_CAs to RECEPTOR_CAs

Dear Patrick I was recently using EvoBind and got the following problem, I think it's because of a new update they have implemented... I don't know if you could guide me on how to solve this error. Thanks a lot

Dear Patrick
I hope this message finds you well, recently I tried to use Colab Notebook of Evobind and I used the test case but when the Colab goes to "Run Prediction", it crashes throwing the following error:
XlaRuntimeError: INTERNAL: RET_CHECK failure (external/org_tensorflow/tensorflow/compiler/xla/service/gpu/gemm_algorithm_picker.cc:309) stream->parent()->GetBlasGemmAlgorithms(stream, &algorithms)

I don't know if it's an execution problem on my part or if it's broken internally

Greetings

Should $IMG be $SINGIMG instead?

Otherwise one gets this

FATAL:   could not open image /dgx/home/userexternal/tgiorgin/EvoBind/python3: failed to retrieve path for /dgx/home/userexternal/tgiorgin/EvoBind/python3: lstat /dgx/home/userexternal/tgiorgin/EvoBind/python3: no such file or directory

because "python" is taken as the image name. (This may be the same issue as #2 )