Prediction of a small dimer with a ligand

[Luke-ebbis]

Description

I am attempting a prediction of the enolase dimer with its ligand 2PG by running these commands:

$ mkdir data
$ wget -O "data/2PG.sdf" "https://files.rcsb.org/ligands/download/2PG_ideal.sdf"
$ python neuralplexer-inference --task=batched_structure_sampling --model-checkpoint results/dependencies/neuralplexer/data/neuralplexermodels_downstream_datasets_predictions/models/complex_structure_prediction.ckpt --out-path results/data/enolase-jobs/enolase-dimer-job1/ --input-receptor 'AVSKVYARSVYDSRGNPTVEVELTTEKGVFRSIVPSGAATGVHEALEMRDGDKSKWMGKGVLHAVKNVNDVIAPAFVKANIDVKDQKAVDDFLISLDGTANKSKLGANAILGVSLAASRAAAAEKNVPLYKHLADLSKSKTSPYVLPVPFLNVLNGGSHAGGALALQEFMIAPTGAKTFAEALRIGSEVYHNLKSLTKKRYGASAGNVGDEGGVAPNIQTAEEALDLIVDAIKAAGHDGKVKIGLDCASSEFFKDGKYDLDFKNPNSDKSKWLTGPQLADLYHSLMKRYPIVSIEDPFAEDDWEAWSHFFKTAGIQIVADDLTVTNPKRIATAIEKKAADALLLKVNQIGTLSESIKAAQDSFAAGWGVMVSHRSGETEDTFIADLVVGLRTGQIKTGAPARSERLAKLNQLLRIEEELGDNAVFAGENFHHGDKL|AVSKVYARSVYDSRGNPTVEVELTTEKGVFRSIVPSGAATGVHEALEMRDGDKSKWMGKGVLHAVKNVNDVIAPAFVKANIDVKDQKAVDDFLISLDGTANKSKLGANAILGVSLAASRAAAAEKNVPLYKHLADLSKSKTSPYVLPVPFLNVLNGGSHAGGALALQEFMIAPTGAKTFAEALRIGSEVYHNLKSLTKKRYGASAGNVGDEGGVAPNIQTAEEALDLIVDAIKAAGHDGKVKIGLDCASSEFFKDGKYDLDFKNPNSDKSKWLTGPQLADLYHSLMKRYPIVSIEDPFAEDDWEAWSHFFKTAGIQIVADDLTVTNPKRIATAIEKKAADALLLKVNQIGTLSESIKAAQDSFAAGWGVMVSHRSGETEDTFIADLVVGLRTGQIKTGAPARSERLAKLNQLLRIEEELGDNAVFAGENFHHGDKL' --input-ligand 'data/2PG.sdf|data/2PG.sdf' --sampler langevin_simulated_annealing --n-samples 20 --chunk-size 10 --num-steps=100 --cuda

I am running these commands using a docker image build from the last main branch.

I obtain a structure where the two dimer subunits show structural clashing that is extremely unrealistic (Table 1). The subunits show an almost complete overlap as can be seen in the contact map (Figure 1).

Table 1: Number of steric clashes

structure	clashes	clashes_percentage	frameid
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	826	13.098636219473518	0
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	858	13.606089438629876	1
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	1110	17.602283539486205	2
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	814	12.908341262289882	3
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	824	13.066920393276247	4
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	883	14.002537266095782	5
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	973	15.429749444973043	6
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	965	15.302886140183952	7
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	1106	17.53885188709166	8
results/data/enolase-jobs/enolase-dimer-job3/prot_all.pdb	1099	17.427846495401205	9

Figure 1: Contact map between the subunits at 15Å between the C$\alpha$ atoms

I feel that there is an issue with the generation process, and that the model does not advance beyond the initialization stage. Am I using the correct parameters for prediction? Can someone suggest better parameters?

[spyda90]

Hi, are you able to download them from here? https://1drv.ms/u/s!AjjQVuMYxf5qgZkURaFW0fbrJqKECA?e=4ICiuD

[Luke-ebbis]

Yes, I will check it and report on the issue

Op di 16 jul 2024 om 12:56 schreef spyda90 @.***>:

Hi, are you able to download them from here? https://1drv.ms/u/s!AjjQVuMYxf5qgZkURaFW0fbrJqKECA?e=4ICiuD

— Reply to this email directly, view it on GitHub https://github.com/zrqiao/NeuralPLexer/issues/49#issuecomment-2230605065, or unsubscribe https://github.com/notifications/unsubscribe-auth/AWVTJMAKBU37EIPW7J6EYDTZMT34ZAVCNFSM6AAAAABJQI4BC6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEMZQGYYDKMBWGU . You are receiving this because you authored the thread.Message ID: @.***>