2
$\begingroup$

So, my runs have finished and got 5 final models. I predicted TM-score of one model to understand the result. The output of "predict-tmscore.sh" for "final_1.pdb" is:

1 models read from PDB file
1 models read from PDB file
nres is 377
neff is 15423
dpscore is -300.599030
dpscorenorm is -0.082424
Probability of each TM-score bin (cols 1-10), estimated TM-score (col 11) and prob TM > 0.5 (col 12) written to sequence.ema.txt
Estimated TM-score of ../final_1.pdb is 0.5

could you please help me understand what does "nres", "neff", "dpscore", and "dpscorenorm" below mean:

    nres is 377 
neff is 15423 
dpscore is -300.599030 
dpscorenorm is    -0.082424

and it also generates a "sequence.ema.txt" file with below value:

2.0308916646172293e-05 2.0682286958617624e-06 0.04564645141363144 0.14857041835784912 0.27177685499191284 0.31814685463905334 0.19150513410568237 0.022412151098251343 0.001872337656095624 4.743059980683029e-05 0.5036169363265117 0.5339840054512024

Please let me know what these above values signify?

And,I would really appreciate your comments on what does "rawdistpred.1" file mean? There are total 9 such files. When predicitng TM-score, do I need to use "rawdistpred.1" file for all final models or I can use any "rawdistpred.*" file?

thanks!

$\endgroup$
1
  • $\begingroup$ Excellent question and answers on two core structure prediction programs $\endgroup$ – M__ Sep 24 '19 at 10:50
2
$\begingroup$

This is described briefly in the paper. The four quantities are inputs to the estimator of model accuracy:

  • nres is the sequence length.
  • neff is an effective sequence count, or how deep the alignment was.
  • dpscore is how well the final model matches the initial distance prediction.
  • dpscorenorm is a normalised version of the above.

Don't worry too much about directly interpreting these. They are only given for completion. The important line is Estimated TM-score of ../final_1.pdb is 0.5, indicating that DMPfold just about thinks this is a decent model.

The "sequence.ema.txt" is more raw output giving the probability of the TM-score being in each bin. "rawdistpred.1" is the distance predictions from the first iteration. Always use "rawdistpred.1" as input to the TM-score predictor as this is how it was trained.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.