Average conservation of 1D structure between remote homologues

Contact: Burkhard Rost (rost@EMBL-Heidelberg.de)
Date: September, 1996

• Question:
  Secondary structure and residue solvent accessbility are conserved between homologues proteins (Rost & Sander, 1994).
  Are these aspects of 1D structure also conserved between remote homologues (i.e. proteins with < 25% pairwise sequence identity)?
• Method:
  The levels of conservation are compiled based on structural alignment taken from the FSSP database. The dataset comprised all remote homologues to a set of 142 structure-unique proteins (i.e. no pair in the set of 142 had a similar structure).
  
• Result and conclusion:
  Although for many remote homologues, neither secondary structure, nor residue solvent accessibility appeared conserved, on average these aspects of 1D structure were found to be conserved. The level of conservation fall off more clearly for solvent accessibility.
  This result is empirical basis for the success of prediction-based threading, i.e., the attempt to recognise similar folds (fold recognition) by aligning 1D structure predictions and/or sequences.
• Figure caption:
  The conservation of 1D structure between remote homologues (<25% pairwise sequence identity) of known structure extends significantly beyond the level at which sequence similarity is clearly recognisable. The histograms show the identity in (a) secondary structure (helix, strand, loop) and (b) relative solvent accessibility (buried, exposed) between pairs of homologous 3D structure averaged over all pairs with a given pairwise sequence identity. Structurally aligned pairs were taken from FSSP (Holm & Sander, 1994). The numbers of pairs used for computing averages is given in the lowest panel (c). The average conservation of secondary structure was 69% (overall three-state identity, one standard deviation was 13%); the average conservation of relative solvent accessibility 65% (overall two-state identity, one standard deviation was 10%).