search

YoshitakaMo / localcolabfold

ColabFold on your local PC
MIT License
554 stars 129 forks source link

AlphaFold Structure Generation using pdb template #248

Open aashuph16221 opened 4 weeks ago

aashuph16221 commented 4 weeks ago

Hello,

I have installed local ColabFold on my local machine and want to generate a structure using a PDB template. I used the following command:

$ colabfold_batch --template --amber --random-seed 10  fasta_sequence.fasta --local-pdb-path ./1.output/temple/ --custom-template-path ./1.output/temple/ ./1.output/

Despite this, I am not able to obtain an AlphaFold2 structure close to the template structure.

Could you please guide me on whether I might have missed something?

Additionally, when running the log file, I received the following output:

2024-08-15 13:31:18,792 Setting max_seq=508, max_extra_seq=1544
2024-08-15 13:34:38,935 alphafold2_multimer_v3_model_1_seed_010 recycle=0 pLDDT=78.4 pTM=0.759 ipTM=0.743
2024-08-15 13:37:06,168 alphafold2_multimer_v3_model_1_seed_010 recycle=1 pLDDT=82.4 pTM=0.824 ipTM=0.813 tol=1.99
2024-08-15 13:39:33,925 alphafold2_multimer_v3_model_1_seed_010 recycle=2 pLDDT=83.1 pTM=0.832 ipTM=0.82 tol=0.256
2024-08-15 13:39:33,926 alphafold2_multimer_v3_model_1_seed_010 took 491.7s (2 recycles)
2024-08-15 13:42:01,127 alphafold2_multimer_v3_model_2_seed_010 recycle=0 pLDDT=68.3 pTM=0.671 ipTM=0.657
2024-08-15 13:44:28,311 alphafold2_multimer_v3_model_2_seed_010 recycle=1 pLDDT=71.6 pTM=0.734 ipTM=0.72 tol=2.14
2024-08-15 13:46:55,894 alphafold2_multimer_v3_model_2_seed_010 recycle=2 pLDDT=80.1 pTM=0.806 ipTM=0.794 tol=0.926
2024-08-15 13:49:24,934 alphafold2_multimer_v3_model_2_seed_010 recycle=3 pLDDT=81.4 pTM=0.813 ipTM=0.8 tol=0.286
2024-08-15 13:49:24,935 alphafold2_multimer_v3_model_2_seed_010 took 590.2s (3 recycles)
2024-08-15 13:51:52,818 alphafold2_multimer_v3_model_3_seed_010 recycle=0 pLDDT=69.9 pTM=0.68 ipTM=0.664
2024-08-15 13:54:20,322 alphafold2_multimer_v3_model_3_seed_010 recycle=1 pLDDT=75.2 pTM=0.754 ipTM=0.739 tol=2.66
2024-08-15 13:56:46,189 alphafold2_multimer_v3_model_3_seed_010 recycle=2 pLDDT=78.2 pTM=0.785 ipTM=0.771 tol=0.441
2024-08-15 13:56:46,190 alphafold2_multimer_v3_model_3_seed_010 took 440.5s (2 recycles)
2024-08-15 13:59:14,282 alphafold2_multimer_v3_model_4_seed_010 recycle=0 pLDDT=70.7 pTM=0.705 ipTM=0.69
2024-08-15 14:01:43,117 alphafold2_multimer_v3_model_4_seed_010 recycle=1 pLDDT=76.2 pTM=0.78 ipTM=0.768 tol=1.79
2024-08-15 14:04:11,503 alphafold2_multimer_v3_model_4_seed_010 recycle=2 pLDDT=80.2 pTM=0.811 ipTM=0.798 tol=0.392
2024-08-15 14:04:11,504 alphafold2_multimer_v3_model_4_seed_010 took 444.5s (2 recycles)
2024-08-15 14:06:39,911 alphafold2_multimer_v3_model_5_seed_010 recycle=0 pLDDT=80.2 pTM=0.792 ipTM=0.779
2024-08-15 14:09:07,560 alphafold2_multimer_v3_model_5_seed_010 recycle=1 pLDDT=85 pTM=0.854 ipTM=0.843 tol=1.87
2024-08-15 14:11:34,444 alphafold2_multimer_v3_model_5_seed_010 recycle=2 pLDDT=85.1 pTM=0.853 ipTM=0.841 tol=0.328
2024-08-15 14:11:34,445 alphafold2_multimer_v3_model_5_seed_010 took 442.2s (2 recycles)
2024-08-15 14:11:35,234 reranking models by 'multimer' metric
2024-08-15 14:11:38,009 Warning: importing 'simtk.openmm' is deprecated.  Import 'openmm' instead.
2024-08-15 14:13:41,635 Relaxation took 126.4s
2024-08-15 14:13:41,635 rank_001_alphafold2_multimer_v3_model_5_seed_010 pLDDT=85.1 pTM=0.853 ipTM=0.841
2024-08-15 14:15:16,654 Relaxation took 95.0s
2024-08-15 14:15:16,655 rank_002_alphafold2_multimer_v3_model_1_seed_010 pLDDT=83.1 pTM=0.832 ipTM=0.82
2024-08-15 14:17:22,475 Relaxation took 125.8s
2024-08-15 14:17:22,475 rank_003_alphafold2_multimer_v3_model_2_seed_010 pLDDT=81.4 pTM=0.813 ipTM=0.8
2024-08-15 14:18:57,755 Relaxation took 95.3s
2024-08-15 14:18:57,755 rank_004_alphafold2_multimer_v3_model_4_seed_010 pLDDT=80.2 pTM=0.811 ipTM=0.798
2024-08-15 14:21:10,270 Relaxation took 132.5s
2024-08-15 14:21:10,270 rank_005_alphafold2_multimer_v3_model_3_seed_010 pLDDT=78.2 pTM=0.785 ipTM=0.771
2024-08-15 14:21:12,676 Done

Regards,

Aashish

YoshitakaMo commented 4 weeks ago

  1. Did your prediction detect and use templates correctly without --local-pdb-path ./1.output/temple/ --custom-template-path ./1.output/temple/ ?

  2. Does your custom-template-path --custom-template-path ./1.output/temple/ contain appropriate mmCIF files for the input fasta file? Templates cannot be selected if the input sequence does not resemble any template mmCIF files in the directory.

harshagrawal13 commented 3 weeks ago

Hi @YoshitakaMo does this repo support inferring orientation and binding region for multimer from a given template? I want a binder (chain B) to bind chain A in a particular site for which I want to specify a template. Would this work?

YoshitakaMo commented 3 weeks ago

does this repo support inferring orientation and binding region for multimer from a given template?

No. Also, since this Localcolabfold repo is focused solely on assisting the installation of ColabFold functionality on a local machine, the question is not the subject here.

aashuph16221 commented 3 weeks ago

Hi @YoshitakaMo, I have found that the issue is related to the mmCIF file format. I obtained the following scores: pLDDT = 61.4, pTM = 0.608, and ipTM = 0.571. My ipTM score seems low, as I read in the AlphaFold documentation that a score >0.8 is considered good for subunit arrangement. I also tried increasing the number of steps, but it has not improved the score. Am I missing any steps, or do you have any suggestions?

YoshitakaMo commented 3 weeks ago

@aashuph16221 Because this repo only supports database setup, installation, and usage in CUI, questions regarding individual cases are not covered here. Perhaps, you are not using it in the wrong way. In my experience, obtaining MSAs of good quality and quantity is more important than using templates for more accurate complex structure prediction.