window mode +/-1 offset?

[sjanssen2]

foldgrammars: eval gives correct overdangle free enery of -13.30 kcal/mol:

(gapc) t490s x86_64 /Daten/Git/jlab/fold-grammars/Misc/Applications/RNAshapes>./x86_64-linux-gnu/RNAshapes_eval_overdangle -P /home/sjanssen/share/gapc/librna/rna_turner2004.par AAGGGCGUCGUCGCCCCGAGUCGUAGCAGUUGACUACUGU "..(((((....))))).........(((((.....)))))"
Answer: 
( ( ..(((((....))))).........(((((.....))))) , -1330 ) , [][] )

using a longer sequence (with the very same first 40 nucleotides) in window mode returns wrong free energy: -13.70

(gapc) t490s x86_64 /Daten/Git/jlab/fold-grammars/Misc/Applications/RNAshapes>./x86_64-linux-gnu/RNAshapes_mfe_overdangle_window -w 40 -i 20 -P /home/sjanssen/share/gapc/librna/rna_turner2004.par AAGGGCGUCGUCGCCCCGAGUCGUAGCAGUUGACUACUGUUuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu
Answer (0, 40) :
( -1370 , ( ( ..(((((....))))).........(((((.....))))) , [][] ) , 28778.4 ) )
Answer (20, 60) :
( -470 , ( ( ....((((((.....))))))................... , [] ) , 0.0130966 ) )
Answer (40, 80) :
( 0 , ( ( ........................................ , _ ) , 6.38663e-06 ) )
Answer (60, 89) :
( 0 , ( ( ............................. , _ ) , 0.000171326 ) )

also, RNAeval from Vienna package (2.4.17) confirms energy of 13.30:

(base) t490s x86_64 /Daten/Git/jlab/fold-grammars/Misc/Test-Suite/StefanStyle/x86_64-linux-gnu>~/src/ViennaRNA-2.4.17/src/bin/RNAeval -v -d2 -T 37 -P /home/sjanssen//share/gapc/librna/rna_turner2004.par

Use '&' to connect 2 sequences that shall form a complex.
Input sequence (upper or lower case) followed by structure; @ to quit
....,....1....,....2....,....3....,....4....,....5....,....6....,....7....,....8
AAGGGCGUCGUCGCCCCGAGUCGUAGCAGUUGACUACUGU
..(((((....))))).........(((((.....)))))

length = 40
External loop                           :   -80
Interior loop (  3, 16) GC; (  4, 15) GC:  -330
Interior loop (  4, 15) GC; (  5, 14) GC:  -330
Interior loop (  5, 14) GC; (  6, 13) CG:  -340
Interior loop (  6, 13) CG; (  7, 12) GC:  -240
Hairpin  loop (  7, 12) GC              :   400
Interior loop ( 26, 40) GU; ( 27, 39) CG:  -250
Interior loop ( 27, 39) CG; ( 28, 38) AU:  -210
Interior loop ( 28, 38) AU; ( 29, 37) GC:  -210
Interior loop ( 29, 37) GC; ( 30, 36) UA:  -220
Hairpin  loop ( 30, 36) UA              :   480
AAGGGCGUCGUCGCCCCGAGUCGUAGCAGUUGACUACUGU
..(((((....))))).........(((((.....)))))
 energy = -13.30 kcal/mol

Is the window mode broken in gapc?

[sjanssen2]

I think the issue is with dall which calls ext_mismatch_energy from rnalib.c. The later probably runs out of the window sub-sequence for P->mismatchExt[closingBP][lbase][rbase].

[sjanssen2]

I am addressing this issue via PR in gapc repo

[sjanssen2]

last window isn't printed by gapc binary :-/

[sjanssen2]

looks like it makes a difference if product is compiled with --kbacktrace or not!
Issue transfers from pAliKiss to RNAalishapes.
we loose the best solution for window (4, 19)

RNAalishapes>./x86_64-linux-gnu/RNAalishapes_subopt_overdangle_window -i 1  -P /home/sjanssen/share/gapc/librna/rna_turner1999.par -u 1        "G_____CGGGAGAC____CG#G_____CGGGAGACG___CG#G_____CGGGAGACG___CG#" -C 0 -w 15 -e 0.1
Answer (0, 15) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( ........(....). , [] ) , 0.0512456 ) )
Answer (1, 16) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( .......(....).. , [] ) , 0.0512456 ) )
Answer (2, 17) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( ......(....)... , [] ) , 0.0512456 ) )
Answer (3, 18) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( .....(....).... , [] ) , 0.0512456 ) )
Answer (4, 19) :
( ( -86.6667 = energy: -86.6667 + covar.: 0 ) , ( ( ...((....)....) , [] ) , 0.0459917 ) )
Answer (5, 20) :
( ( -296.667 = energy: -296.667 + covar.: 0 ) , ( ( .(((....)....)) , [] ) , 1.38822 ) )

(base) t490s x86_64 /Daten/Git/jlab/fold-grammars/Misc/Applications/RNAalishapes>./x86_64-linux-gnu/RNAalishapes_subopt_overdangle_window -i 1  -P /home/sjanssen/share/gapc/librna/rna_turner1999.par -u 1        `cat ~/eclipse-workspace/librna_change/Debug/mininput.msa | tr "\n" "#"` -C 0 -w 15 -e 0
Answer (0, 15) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( ........(....). , [] ) , 0.0512456 ) )
Answer (1, 16) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( .......(....).. , [] ) , 0.0512456 ) )
Answer (2, 17) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( ......(....)... , [] ) , 0.0512456 ) )
Answer (3, 18) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( .....(....).... , [] ) , 0.0512456 ) )
Answer (4, 19) :
Answer (5, 20) :
( ( -296.667 = energy: -296.667 + covar.: 0 ) , ( ( .(((....)....)) , [] ) , 1.38822 ) )

If compiled without --kbacktrack we find the best solution:

Answer (4, 19) :
( ( -93.3333 = energy: -93.3333 + covar.: 0 ) , ( ( ....(....)..... , [] ) , 0.0512456 ) )

[sjanssen2]

if we aim to look at the full search space by increasing -e 100000, we see that every window lacks candidates that are present a) if compiled without --kbacktrace or if we print the forward res in out_main.cc.

[sjanssen2]

A non alignment, gapc self contained example:

(base) t490s x86_64 /Daten/Git/jlab/gapc_libtest/testdata/regresstest/temp>gapc -p "mfe * pretty" ../../grammar/adpf.gap    --window-mode -t --kbacktrack 
Warning: 
  scoring choice [double] h([double] l)
                          ^
../../grammar/adpf.gap:590.27: Declared role of algebra choice function h (in algebra p_func_sample) does not match autodetected role none(n).

(base) t490s x86_64 /Daten/Git/jlab/gapc_libtest/testdata/regresstest/temp>make -f out.mf CXXFLAGS_EXTRA="-I ../../gapc_filter/"
Including global makefile /home/sjanssen/share/gapc/config_linux-gnu.mf
echo '#include "out.hh"' > out_main.cc
cat /home/sjanssen/include/rtlib/generic_main.cc >> out_main.cc
g++  -I/usr/include  -I/usr/include -I/home/sjanssen/include/ -I/home/sjanssen/include/rtlib -I/home/sjanssen/include/librna  -O3  -std=c++17 -D_XOPEN_SOURCE=500 -MMD -MP -Wall -Wnon-virtual-dtor -Wno-unused-variable -Wno-parentheses -DNDEBUG -DBISONNEW  -I ../../gapc_filter/ out_main.cc -c -o out_main.o 
g++  -I/usr/include  -I/usr/include -I/home/sjanssen/include/ -I/home/sjanssen/include/rtlib -I/home/sjanssen/include/librna  -O3  -std=c++17 -D_XOPEN_SOURCE=500 -MMD -MP -Wall -Wnon-virtual-dtor -Wno-unused-variable -Wno-parentheses -DNDEBUG -DBISONNEW  -I ../../gapc_filter/ out.cc -c -o out.o 
g++  -o out out_main.o out.o string.o  -L/usr/lib/x86_64-linux-gnu  -L/home/sjanssen/lib/ -L/home/sjanssen/lib/rtlib -L/home/sjanssen/lib/librna   -Xlinker -rpath -Xlinker /usr/lib/x86_64-linux-gnu      -Xlinker -rpath -Xlinker /home/sjanssen/lib/ -Xlinker -rpath -Xlinker /home/sjanssen/lib/rtlib -Xlinker -rpath -Xlinker /home/sjanssen/lib/librna    -L/usr/lib/x86_64-linux-gnu -lgsl -lgslcblas -lm -lrna 

(base) t490s x86_64 /Daten/Git/jlab/gapc_libtest/testdata/regresstest/temp>./out ucgaagacuugguccuuuggug -P ../../../librna/vienna/rna_turner1999.par -w 20 -i 1 
Answer (0, 20) :
( -300 , .(((((........))))). )
Answer (1, 21) :
Answer (2, 22) :

[sjanssen2]

iff ext_mismatch_energy always returns the same value, the problem vanishes?!

[sjanssen2]

here comes a really small example containing this error:

Save the following code as adpf.gap, compile via

(base) t490s x86_64 /Daten/Git/jlab/gapc_libtest/testdata/regresstest/temp>gapc -p "mfe * (enum)" ../../grammar/adpf.gap --window-mode -t --kbacktrack && make -f out.mf CXXFLAGS_EXTRA="-I ../../gapc_filter/"

and execute binary via

(base) t490s x86_64 /Daten/Git/jlab/gapc_libtest/testdata/regresstest/temp>./out ucgaag -w 4 -i 1 which will result in the output

Answer (0, 4) :
( -160 , cadd( dlr( <0,0> hl( <0,2> ) <2,2> ) cadd( dlr( <2,2> hl( <2,4> ) <4,4> ) nil( 1 ) ) ) )
Answer (1, 5) :
( -160 , cadd( dlr( <1,1> hl( <1,3> ) <3,3> ) cadd( dlr( <3,3> hl( <3,5> ) <5,5> ) nil( 1 ) ) ) )
Answer (2, 6) :

adpf.gap:

import rna
import adpf_filter

input rna

signature FS (alphabet, comp) {
  comp sadd(Subsequence, comp);
  comp cadd(comp, comp);
  comp dlr(Subsequence, comp, Subsequence);
  comp hl(Subsequence);
  comp nil(void);
  choice [comp] h([comp]);
}

algebra mfe implements FS(alphabet = char, comp = int) {
  int sadd(Subsequence lb, int e) {
    return e;
  }
  int cadd(int x, int y) {
    return x + y;
  }
  int dlr(Subsequence lb, int x, Subsequence rb) {
    return x;
  }
  int hl(Subsequence lb) {
    return fake_energy(lb, lb);
  }
  int nil(void) {
    return 0;
  }
  choice [int] h([int] i) {
    return list(minimum(i));
  }
}

algebra enum auto enum;
algebra count auto count;

grammar fold uses FS(axiom = struct) {
   struct = sadd(REGION with minsize(2) with maxsize(2), struct)
          | cadd(dangle, struct)
          | nil(EMPTY) 
	  # h ;

   dangle = dlr(LOC, hl( REGION with minsize(2) with maxsize(2)), LOC) 
	  # h;
}

instance ppmfe = fold (  mfe ) ;