Back to "Biological Language Modeling Seminar Topics"

Back to "Protein structure prediction"

Threading

= method for fold recognition by building many rough models of the query protein sequence based on different possible alignments of the sequences of the known and unknown proteins: i.e. try out all alignments by pulling the query sequence through the 3d space of any known structure

Difference to homology modeling:

Homology modeling                                            Threading

first identify homologues                                    try all possible parents

determine optimal alignment                              try many possible alignments

optimize the model                                             evaluate many rough models*

* Evaluation of the rough models:

    - based on the distribution of inter-residue distances for all 20x20 pairs of residue types

    - relate these probabilities to energies using the Maxwell-Boltzmann equation to obtain an energy function that is used to score the models

How do the results differ?