| Mistake | Cause | Result | Fix |
| over-interpretation of sequence similarity | level too low (<30%) too many gaps similarity confused with sequence identity composition bias
| incorrectly inferred homology, hence wrong function or wrong structure | use thresholds considering gaps, similarity and composition bias |
| insufficient description of family | full family not matching local motifs | pattern / homology incorrectly inferred | use profiles of sub-family repeat search with different family members |
| inaccurate function designation in database | errors in original paper wrong annotation in database over-interpretation of homology in database | incorrectly inferred function too detailed predictions for function | check literature compare several homologues compare with level of divergence in family |
| over-interpretation of secondary structure prediction | single residue assignments taken too literally | wasted time incorrectly inferred function | consider mainly segment level |
| over-interpretation of threading | trendy false prediction prediction not reliable | wrong or partially wrong 3D model | use controls appropriate to particular threading method check functional residues |
| over-interpretation of 3D model | 3D model built on shaky alignment information, or unreliable structures large loops modelled wrong choice of modelling and/or checking tools | wrong interpretation of location of residue in 3D structure wrong prediction of structure-function relationship | do more careful alignment take into account reliability index for homology models low level similarity => coarse-grained tools high level similarity => full atom description tools
|