I want to search *structurally* within a not-so-well-characterized set of proteins from a genome. I am looking for specific secondary structure patterns or domains, indicative of possible ancient, deep relationships with known domains. I cannot hope or rely on significant sequence similarity. What I need is something akin to "Please find, among these 5000 proteins, the ones who could have, say, three alpha-helices plus two beta-strands plus another two alpha-helices". Once I have this set, I could then look at it in more detail. Is there a more or less trusted server or software or method to do something like that? Thanks a lot!