Kevin Karplus 24 May 1998 t55 is clearly homologous to 1esl, 1rtm1, 1lit, ... Since 1esl gets the highest score, this is probably the one to work with, though 1rtm1 is very nearly as good. 1hlj, which does not appear in fssp's TABLE2 scores slightly higher with the target model, so may be worth looking at. Since we know that this C-type lectin recognizes Galactose, it might be worth checking to see if any of the lectins in PDB do also. Active site alignment is clearly the goal here. 26 May 1998 Kevin Karplus 1esl-t55-fssp-global alignment looks good, preserves cystine bonds. 16 July 1998 Kevin Karplus Blast and double-blast get 1kja and 1esl as top hits, with very strong scores. Note: 1kja is a theoretical model, so should be ignored. Although we have a strong homology, the alignment does not look completely trivial. I'm going to re-create the set of alignments with new SAM and t55.t98_5. These alignments all seem to be quite similar---probably any one would do. 1esl-t55-vit = 1esl-t55-post = 1esl-t55-global t55-1esl-post = t55-1esl-global = t55-1esl-vit shift measure 1esl-t55-fssp 1esl-t55-const 0.965939 t55-1esl-vit t55-1esl-joint 0.945455 t55-1esl-joint 1esl-t55-const 0.945263 1esl-t55-joint 1esl-t55-const 0.926222 1esl-t55-joint 1esl-t55-vit 0.924434 t55-1esl-vit 1esl-t55-const 0.919828 t55-1esl-vit 1esl-t55-fssp 0.918844 t55-1esl-joint 1esl-t55-vit 0.918590 t55-1esl-joint 1esl-t55-joint 0.910314 ... It looks like 1esl-t55-fssp is a good starting alignment, having good shift scores with respect to essentially all the other alignments. Its worst agreement is with 1esl-t55-joint (0.90), so these two alignments should be compared. Overall worst agreement is between 1esl-t55-vit and t55-1esl-vit, so maybe they should be compared as well. 18 July 1998 I looked at the t55.t98_5_sorted alignment to see what 3D alignments were most promising. It turne dout that different parts of the alignment had different best templates (based on residue ID): residues template 1-30 1esl 30-50 none had great alignment 50-75 1hli OOPS this is a theoretical model, continue 1esl 76-120 1hup Better split 1-103 1esl 104-120 1hup I'll have to do joint models for each of these, and see whether we want to provide 1 or 2 segments to our predictions. There is a fssp alignment between 1esl and 1rtm1, and between 1rtm1 and 1hup. This pair of alignments is consistent with the 1esl-1hup alignment from t55.t98_5. Note: using FSSP alignment 1esl-1rtm1-1hup basically gets the same alignment for t55-1esl and t55-1hup at the end, but sidechains may be better packed in 1hup at the end and in 1esl at the beginning. The 1esl-t55-const-global alignment agrees with the t55.t98_5 alignment except in slight shifts of gaps and insertions. 21 July 1998 Christian I don't think that using 1hup as a template for the last 16 or so residues of t55 is a good idea. While this region in 1hup and 1esl has the same topology, its structure is quite different owing to an immediately preceding beta sheet that exists in 1hup but not 1esl. I don't think the structure piece from 1hup is compatible with the rest of 1esl. My alignment for t55 using 1esl as a template is 1esl-t55-fssp-global.cbarrett-hand.pw.a2m. Wed Jul 22 18:12:54 PDT 1998 OK, let's got with 1esl. I modified Christian's alignment slightly to get 1esl-t55-kevin-hand.a2m. It may not be any better, but I like the placement of the gaps a bit better. Christian, do you want to argue for or against any of the small changes I made? 23 July 1998 Draft of results: T0055 is clearly homologous to several existing structures, including 1ESL, 1RTM1, 1LIT, 1HUP, ... The 1ESL homology was strong enough for wu-blast to find. Double-blast found 60 homologous sequences in PDB (though many were theoretical models or duplicates of other sequences). Our method scored 1ESL highest at -134.7, well into the range where we had no false positives on superfamily recognition and the other structures in the same superfamily also scored very well. The first 103 residues had the greatest similarity to 1ESL, and the next 16 to 1HUP. We considered doing a hybrid prediction, cutting and pasting the two, but it seemed simpler just to use 1ESL throughout---the greater identity over 16 residues was unlikely to compensate for the difference in environment. We did a small amount of hand-tweaking of the alignment to place the gaps in convenient places in the 3D strucutre.