For example, if we want to study frequency of a specific mutation (very common codon change) and its structural consequences. From an evolutionary or universal point of view, what is minimum number of samples that is statistically acceptable for such study?
I would suggest to obtain proteins from each of the 14 SCOP families, The families differ in the number of proteins they contain, so get a number of samples from each family in proportion to its size. But note that the PDB itself is inherently a very biased dataset: it is biased in favor of proteins that are readily available and crystallizable. Proteins that won't crystallize are of course not represented at all in the PDB.
F-U: I just remembered that the PDB group has performed some structural clustering of all the proteins deposited in the PDB, and in so doing has arrived at what they call a nonredundant protein structure dataset. I just don't know how many structural equivalence classes they have now; you may contact them for this info. Such a nonredundant dataset would a very good sample on which to "perform a study on a general structural characteristic," if I may use your own words.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Proteins