Page 1 of 1
D-I-TASSER results
Posted: Fri Jun 06, 2025 7:03 am
by Abd_Compbio
I ran a job on the D-I-TASSER server and analyzed the results with MolProbity. Compared to Alphafold3, I expected results to be more aligned with MolProbity score highlighting structural inconsistency, including steric clashes, bond angles, Ramachandran outliers,
[/img] etc. Alphafold3 resulted in a percentile of 94, and D-I-TASSER results are on the 3rd percentile with very bad errors. I would like to know more about this in detail. Is there a problem with a server? Or problems with results? I have attached an image of my results, also of both alpha bold and DITASSER based:
https://drive.google.com/drive/folders/ ... sp=sharing
Re: D-I-TASSER results
Posted: Sun Jun 08, 2025 12:04 pm
by junh
Could you tell us the D-I-TASSER job ID. Thanks.
Re: D-I-TASSER results
Posted: Sun Jun 08, 2025 3:33 pm
by Abd_Compbio
junh wrote: ↑Sun Jun 08, 2025 12:04 pm
Could you tell us the D-I-TASSER job ID. Thanks.
(JOB ID: D1250)
This was job id,
Thanx!
Re: D-I-TASSER results
Posted: Mon Jun 16, 2025 8:36 am
by MableMilton
That's an interesting observation! While I appreciate the expectation of closer alignment with MolProbity scores, AlphaFold3's advancements might prioritize overall accuracy over individual structural consistency metrics in certain cases. Perhaps the underlying algorithm is making different trade-offs. Further exploration is needed to understand when this divergence occurs and what it might reveal about the model's limitations and the data it was trained on, which is not always reflective of ideal in vivo situations.
Re: D-I-TASSER results
Posted: Mon Jun 16, 2025 1:19 pm
by junh
Abd_Compbio wrote: ↑Fri Jun 06, 2025 7:03 am
I ran a job on the D-I-TASSER server and analyzed the results with MolProbity. Compared to Alphafold3, I expected results to be more aligned with MolProbity score highlighting structural inconsistency, including steric clashes, bond angles, Ramachandran outliers,
[/img] etc. Alphafold3 resulted in a percentile of 94, and D-I-TASSER results are on the 3rd percentile with very bad errors. I would like to know more about this in detail. Is there a problem with a server? Or problems with results? I have attached an image of my results, also of both alpha bold and DITASSER based:
https://drive.google.com/drive/folders/ ... sp=sharing
The FG-MD program is not run correctly in D-I-TASSER. Now we fix it. Please try it again. Or you can directly use the FG-MD server (
https://zhanggroup.org/FG-MD) to remove the clashes.
Re: D-I-TASSER results
Posted: Mon Jun 16, 2025 2:26 pm
by Abd_Compbio
junh wrote: ↑Mon Jun 16, 2025 1:19 pm
Abd_Compbio wrote: ↑Fri Jun 06, 2025 7:03 am
I ran a job on the D-I-TASSER server and analyzed the results with MolProbity. Compared to Alphafold3, I expected results to be more aligned with MolProbity score highlighting structural inconsistency, including steric clashes, bond angles, Ramachandran outliers,
[/img] etc. Alphafold3 resulted in a percentile of 94, and D-I-TASSER results are on the 3rd percentile with very bad errors. I would like to know more about this in detail. Is there a problem with a server? Or problems with results? I have attached an image of my results, also of both alpha bold and DITASSER based:
https://drive.google.com/drive/folders/ ... sp=sharing
The FG-MD program is not run correctly in D-I-TASSER. Now we fix it. Please try it again. Or you can directly use the FG-MD server (
https://zhanggroup.org/FG-MD) to remove the clashes.
I was wondering does these errors persist if i have downloaded DITASSER suit earlier? or are these issues with server only?
I have tried FG-MD using server but the results are not that upto mark but atleast better than alpha fold. Score increased from 67 to 76 for molprobity.
Re: D-I-TASSER results
Posted: Wed Jun 18, 2025 7:48 am
by junh
Abd_Compbio wrote: ↑Mon Jun 16, 2025 2:26 pm
junh wrote: ↑Mon Jun 16, 2025 1:19 pm
Abd_Compbio wrote: ↑Fri Jun 06, 2025 7:03 am
I ran a job on the D-I-TASSER server and analyzed the results with MolProbity. Compared to Alphafold3, I expected results to be more aligned with MolProbity score highlighting structural inconsistency, including steric clashes, bond angles, Ramachandran outliers,
[/img] etc. Alphafold3 resulted in a percentile of 94, and D-I-TASSER results are on the 3rd percentile with very bad errors. I would like to know more about this in detail. Is there a problem with a server? Or problems with results? I have attached an image of my results, also of both alpha bold and DITASSER based:
https://drive.google.com/drive/folders/ ... sp=sharing
The FG-MD program is not run correctly in D-I-TASSER. Now we fix it. Please try it again. Or you can directly use the FG-MD server (
https://zhanggroup.org/FG-MD) to remove the clashes.
I was wondering does these errors persist if i have downloaded DITASSER suit earlier? or are these issues with server only?
I have tried FG-MD using server but the results are not that upto mark but atleast better than alpha fold. Score increased from 67 to 76 for molprobity.
Does the clash number is not suitable?
Re: D-I-TASSER results
Posted: Wed Jun 18, 2025 10:41 am
by Abd_Compbio
junh wrote: ↑Wed Jun 18, 2025 7:48 am
Abd_Compbio wrote: ↑Mon Jun 16, 2025 2:26 pm
junh wrote: ↑Mon Jun 16, 2025 1:19 pm
The FG-MD program is not run correctly in D-I-TASSER. Now we fix it. Please try it again. Or you can directly use the FG-MD server (
https://zhanggroup.org/FG-MD) to remove the clashes.
I was wondering does these errors persist if i have downloaded DITASSER suit earlier? or are these issues with server only?
I have tried FG-MD using server but the results are not that upto mark but atleast better than alpha fold. Score increased from 67 to 76 for molprobity.
Does the clash number is not suitable?
Clash number got resolved but while doing this, it created errors in Ramachandran outliers compared to earlier structure.