to answer Dr. Kister's question, the answer is no.

Q:if we add HB info, can we detect a position to cut?

A:Nop

===================
in 1bfg, these three sets of residues all have correct H-Bonds

8 5 9 10: 96,94,63,65,106,104,139,141: fit! yes
1 4 10 9: 23,21,53,55,141,139,104,106: fit! yes
5 8 4 1: 65,63,94,96,55,53,21,23: fit! yes
-----------------------------------------------------
1swu "A"
8 7 1 2: 130,128,106,108,23,21,29,31: fit! yes
which is not same as Martin's
Martin's is 4 3 5 6: 56,54,43,45,79,77,90,92
-----------------------------------------------------
1efc A
22 19 21 20: 363,361,302,304,339,337,330,332: fit! no
20 21 23 24: 332,330,337,339,376,374,386,388: fit! HB yes, position no.
24 23 19 22: 388,386,374,376,306,304,360,362: fit! yes
---------: fit!
21 20 22 19: 339,337,330,332,362,360,304,306: fit! yes
19 22 24 23: 306,304,359,361,388,386,374,376: fit! HB yes, position no.
19 22 24 23: 306,304,360,362,388,386,374,376: fit! yes

Martin
select 362,360,304,306,339,337,330,332; color yellow;

save direction info in table also

For example, we consider in the paper the motif
>
> 1 3 7 5
> 2 9 8 4 6
>
> the question: Does the rule of handedness allow to change the order of
> strand like
>
> 3 1 7 5
> 2 9 8 4 6
>
-- The answer depends on the durection of strands, specificalle because
you often divide one extended piece ("strand" in a sense of topology)
into two beta-strands ("strand" in a sense of beta-sheet). To avoid
this below? I will mark strands going FROM US by +, and TO US by *,
and - THE MAIN - I WILL DRAW CONNECTION between strans:
by double-line when it is on "our" side of the sheet,
by sindle-line when it is on the opposite side,
by : when it is in the middle, i.e.,an extended piece leaves onr sheet
and goes to another).
   Thus, there are two possibilities for strands 1, 2, 3:
1) All beta-strands are full-lenth strands (from one side of
beta-sheet to another)
---------------------               ---------------------
  1   3   7   5                       1   3   7   5     
  *   *                               +   +             
  /                 or, the same,   /              
  2   9   8   4   6                   2   9   8   4   6 
  +                                   *                 
---------------------               ---------------------
Strands 1, 2, 3 form a right-handed turn (anti-clockwise turn of the
chain when it comes to us - follow the connections; we look
perpendicularly to the strands, along axis of the turn formed by
strands 1, 2, 3). Left and right figures have opposite direstion of
the chain, but the same handedness (follow the connections!).
  In this case, the change that you propose (1 <=> 3, without changing
dirsctions of the strands) gives:
---------------------               ---------------------
  3   1   7   5                       3   1   7   5     
  *   *                               +   +             
   //                or, the same,    //              
  2   9   8   4   6                   2   9   8   4   6 
  +                                   *                 
---------------------               ---------------------
Handedness changed! - follow the connections!
"Rule of handedness" forbids this change...

2) Beta-strand in the middle of the sheet is a full-lenth strand (from
one side of beta-sheet to another), and two beta-strands at the edge
are "half-srands" (I wrote you about that)
---------------------               ---------------------
  1   3   7   5                       1   3   7   5     
  *   +                               +   *             
  : /                 or, the same,   : /               
  2   9   8   4   6                   2   9   8   4   6 
  *                                   +                 
---------------------               ---------------------
In this case "handedness" is not defined, sinse we have not a full
(~360deg) turn from strand 1 to 3, but half-a-turn (~180deg).
   Actually, in this case - as I recommended you - you should draw
---------------------               ---------------------
  1a  2   6   4                       1a  2   6   4     
  *   +                               +   *             
  : /                 or, the same,   : /               
  1b  8   7   3   5                   2b  8   7   3   5 
  *                                   +                 
---------------------               ---------------------

  In this case, the change that you propose (1 <=> 3, without changing
dirsctions of the strands) gives:
---------------------               ---------------------
  3   1   7   5                       3   1   7   5     
  *   +                               +   *             
  : //                or, the same,   : //               
  2   9   8   4   6                   2   9   8   4   6 
  *                                   +                 
---------------------               ---------------------
and "handedness" is not defined again.
"Rule of handedness" is silent in this case...

bend

1b34 is the one actually been bend, but can't be detected by distance method.
even increase error to 2.00
PS: increase error to 0.5 can detect sheets in sandwich

residues on barrel

use protein explore and use following code in chime.

1bia
select 289,287,276,278,300,298,309,311; color red;
1bgf
select 65,63,94,96,55,53,21,23; color red;
1efc
select 362,360,304,306,339,337,330,332; color red;
select 331,329,337,339,376,374,386,388:A;color green;

the rule seems like...

oops,
although the rules seems perfect,
after I done more analysis, it seems broke.

when I increase the error to 3A,
although I can hit all the 4 pair strands,
I also hit sandwich.

you'll hit sandwich when error's about 2A.

it seems like it will try to hit a sheet more then four strands.

but the question is, why the "bend" appears the way as my rule?

maybe we don't need 8 residues for barrel

maybe we only need 4 residues for the matrix
because
1bia
11 10 12 13: 289 287 276 278 300 298 309 311: fit!
12 13 11 10: 302 300 307 309 287 285 278 280: fit!

distance matrix tree for barrels by matlab

import 051107:distmatrixforbarrel.xls:Sheet6 into matlab by copy through clipboard first
then execute following program in matlab

A=clipboarddata;
for i=1:27,
    mat1=A((i-1)*8+1:i*8,:);
    val=[];
    for k=1:8,
        val=[val mat1(k,:)];
    end
    mat(i,:)=val;
end
P=pdist(mat,'euclid');
z1=linkage(P,'average');
h101=figure(1);
//close this window first then execute below;
dendrogram(z1);
l={'1bia' '1pdr' '1g3p' '1b34' '1whi' '1sty' '1bfg' '1i8d' '1efc' '1flm' '1fmt' '1eax' '1bco' '1e79' '1daz' '1ile' '2eng' '1h9d' '1mai' '1ieg' '1pkm' '1ik9' '1swu' '2cpl' '1jnp' '1c39' '1bia'};
hx=get(get(h101,'CurrentAxes'),'XTickLabel');
lab=[];
for i=1:size(hx,1)
  hy(i)=str2num(hx(i,:));
  lab=[lab,l(hy(i))];
end
set(get(h101,'CurrentAxes'),'XTickLabel',lab);

try contact & graph

1ph3A

77
84 86 88
12 14 16
38 42 45
62 66 68 70
93 96


1bia 271-317
307 309 311
298 300 302
285 287 289
276 278 280
314
8 varianes

1pdr
465 467 469 471
476 478 480
494 491 489
506 510 512 515
545 543 541 539

tested three structures from three different fold,
I can't find anyone match the graph you want me to draw you. (graph2.doc)

contact residues in barrel

1bia A
init 278 285 289 300 302 309

307/  309= 311$
+-= x*%+- %=


298-  300+
%=\^ *%=/\^?,&

280# 283x  285* 286, 287? 289%=$ %= %+=& +& +& +-^& 


276&  278%
*+\^?, x*+-=^$
 
314^
&%-\+&

ans: 276 278 280 285 289 298 300 302 309 311 314

seems like the bended strand is caused by a missed residue side chain (ex:Glu)
all 2x2 matrix are hydrophobic / side chains are inside.

web tool about motifs in sandwich or barrel

Website: SMoS
Paper: http://peds.oxfordjournals.org/cgi/reprint/16/11/791

interlock in barrel?
http://www.ncbs.res.in/~faculty/mini/SMoS/02.02.092/02.02.092.htm

core of barrel should be four strands?http://www.ncbs.res.in/~faculty/mini/SMoS/02.02.092/02.02.092.htm
http://www.ncbs.res.in/~faculty/mini/SMoS/02.02.094/02.02.094.htm
or, strands face to face?http://www.ncbs.res.in/~faculty/mini/SMoS/02.02.097/02.02.097.htm

idea about strandon prediction

motif in this paper might be the loop within strandon.

foldminer

I tried Fold Miner, it's pretty interesting,
it use statistics to know what SSEs are more importent then other SSEs in a group.
the group can be any node in SCOP database.
it can automatically identify the interlock strands for 1bww.
PS: the statistics is based on whole SCOP database, it takes about 15 mins for one query.

tool
http://motif.stanford.edu/foldminer/
paper
http://www.pubmedcentral.gov/picrender.fcgi?artid=441527&blobtype=pdf