13 July 1998 Kevin Karplus Clear homology to 2por (even wu-blast finds it), as stated by submitters. Double-blast likes 1gfn best, which has fssp rep 2omf, a fairly similar porin. t70.t98_6 gets 1pho as the top scorer, which has fssp rep 2omf. t70-t98 gets the best score for library model 2omf (with 2por as second best). The sum models have the same order. The alignments drift away from t70---t70.remote_2 is ok, but t70.remote_3 is trashed, t70.t98_3 is ok, but t70.t98_4 has lost the target sequence. Using t70.t98_3 gets 1gfn (-21.6) as top-scorer (fssp rep 2omf). The 2omf alignments look fine (I've looked at 2omf-t70 and 2omf-t70-constr-global), but will need some tweaking to make the gaps easier to close. Perhaps aligning 1gfn to the 2omf constrained model will give a better starting point. The 2por alignments also look good, with similar tweaking needed. Both sets of alignments have conserved residues in stripes across the beta sheets, locking the alignment firmly in place---the only tweaking possible is at the ends of the strands. The 1gfn alignment is similarly fairly good, though I think it may have more insertions in awkward places than the others. All the alignments will have to be carefully compared (perhaps by propagating alignment information through the fssp alignments and using measure_shift to show the resulting differences). The secondary structure prediction is clearly wrong in places, with helices predicted where the real porins must have beta strands. 14 July 1998 Kevin Karplus The tricky part is not going to be the beta strands, but guessing what happens inside the porin. Mon Aug 3 14:15:54 PDT 1998 Hand-edited 3 of the alginments: 2omf-t70-hand1: 70 residues conserved 2por-t70-hand1: 63 residues conserved t70-1gfn-hand1: 63 residues conserved 2omf-t70-hand2: 74 residues conserved, alignment matches 1gfn on most conserved residues Actually, the betas are pretty tricky, because there are many alignments of similar quality. I've tried to line up the conserved residues in adjacent positions in the sheets, and leave the major inserts or deletes to the edges of the barrel (or the center). I think I'll go with the 2omf-t70-hand2 alignment---it looks a little better than the others. I definitely like the 1gfn and 2omf alignments better than 2por. The secondary structure prediction is poor---the t98_2 may be slightly better than the t98_3 one, because it uses an alignment that doesn't put any gaps in the guide sequence. The number of errors in both is similar. Draft of comments on results: Even blast recognizes T0070 as a porin, scoring 2por first. Double-blast moves 1gfn to the top, using as an intermediate GenPept sequence BCUOPCP1_1. We had to drop back to the third iteration of the target alignment to get a model that distiguished well among the closely homologous porins. This model also scored 1gfn first, with close homologs 1mpf, 1opf, and 2omf close behind. The 2omf model scored highest of the library models, with 2por close behind. We chose the 2omf model, but had to modify the alignment by hand to get good conservation in stripes across the beta sheets.