2
$\begingroup$

I am absolutely new to Structural Bioinformatics (I only started last week). I am working on the secondary structure assignment/prediction (actually I am not sure) of proteins using ML.

When it comes to protein analysis, most of the books or research papers talk about protein structure Prediction:

https://bip.weizmann.ac.il/course/structbioinfo/pdf/SecondaryStructurePrediction_WIS.pdf

Only a handful of books or articles talk about protein structure Assignments:

https://www.biorxiv.org/content/10.1101/2021.02.02.429329v1

Can anyone tell me what is the basic difference between Assignment and Prediction?

Kindly explain with examples.

$\endgroup$
1
  • $\begingroup$ Do you have any examples for the use of Protein Structure Assignment? This paper uses the term in the title but they don't particularly distinguish the terminology. $\endgroup$ Jul 5 at 5:40
7
$\begingroup$

To make it crystal clear (and make than pun):

  • Assignment: a 3D structure is known and the residues are assigned a secondary structure
  • Prediction: an algorithm predicts from linear (primary) sequence what the secondary structure may be —often incorrectly. This can be homology based, covariance, deep learning etc. But most commonly just the vector norm of window residue composition, e.g. proline is never in a helix and helices and sheets have different periodicity (two for a sheet and three and a bit for a helix).
$\endgroup$
2
  • $\begingroup$ What do you mean by 3D structure? Is it a Secondary or Tertiary structure? $\endgroup$
    – user366312
    Aug 22 at 2:05
  • $\begingroup$ Tertiary/quaternary structure —which may be of only part of the protein and with gaps due to technical limitations. But I mean specifically a list of atomic coordinates (eg. [PDB file](en.m.wikipedia.org/wiki/Protein_Data_Bank_(file_format) )) of that chain, possibly as an oligomer, with a ligands or other chains as opposed to say just the name of the fold, e.g. Rossmann fold. $\endgroup$ Aug 22 at 6:42
0
$\begingroup$

Answer from @jan-wilamowski, converted from comment:

That second paper seems to distinguish between the known data that is used for ML training (i.e. amino acid sequences are assigned some secondary structure) and the attempt to replicate this natural assignment by an "artificial" prediction.

$\endgroup$

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.