errata-3.2

mfold

This is the errata file for version 3.2

1. 8/26/05

The new version of filter-sort for mfold 3.2 had an swk script
error. SIZE is now defined properly using "-v SIZE=$SIZE". Sorting is
performed correctly and tmp2-$1 is not needed.

Old versus new:
6a7
> # 8/26/05 - Redone using awk in July 05 and now corrected.
25,26c26,28
< head -1 $1 > tmp1-$1
< tail +2 $1|tr -s "\011 " "\011\011"|awk '{if($2 >= $SIZE) print $0 }' > tmp2-$1
---
> head -1 $1|tr -s "\011 " "\011\011" > tmp1-$1
> tail +2 $1|tr -s "\011 " "\011\011"|\
>     awk -v SIZE=$SIZE '{if($1 == 1 || $2 >= SIZE) print $0}'|sort -n +4 >> tmp1-$1
28,30d29
< # Now sort and replace
< 
< sort -r +4 tmp2-$1  >> tmp1-$1
32d30
< rm -f tmp2-$1
34a33
> 

2. 9/11/05

The old way of building a heap for automatic traceback worked along
rows. The column would jump by cntrl(9) (window parameter) when an
acceptable base pair was found. Otherwise, the next column would be
used. This was always a rather poor "ad hoc" method. It is also
possible that the minimum energy folding could be missed. For this
reason, traceback initiation base pairs are chosen along
"anti-diagonals" i+j = constant = diag. The first base pair (either on
first row (i=1) or last column (j=n) is always accepted if ene is <=
crit. If (i,j) is accepted, the next base pair to be accepted along an
anti-diagonal, (i',j'), has the property that ene(i',j') <= crit and
ene(i',j') != ene(i,j). In terms of the .plot file, it means that the
outermost base pair is chosen for each helix that is within the energy
increment. 

New is >
Old is <
59,78c59,72
<       do diag = 3,2*n-1
<          if (diag.le.n+1) then
<             i = 1
<             j = diag - i
<          else
<             i = diag - n
<             j = diag - i
<          endif
<          oldene = infinity
<          do while (j.gt.i)
<             if (ene(i,j).le.crit.and.ene(i,j).ne.oldene) then
<                if (num.eq.sortmax) then
<                   err = 31
<                   call errmsg(err,hstnum(i),hstnum(j))
<                   goto 10
<                endif
<                num =  num + 1
<                oldene = ene(i,j)
<                heapi(num) = i
<                heapj(num) = j
---
>       i = 1
>       j = 2
>       do while (i.lt.n)
>          if (ene(i,j).le.crit) then
>             if (num.eq.sortmax) then
>                err = 31
>                call errmsg(err,hstnum(i),hstnum(j))
>                goto 10
>             endif
>             num =  num + 1
>             heapi(num) = i
>             heapj(num) = j
>             j = j + cntrl(9) + 1
>             if (j.gt.n) then
80,81c74,78
<                j = j - 1
<             else
---
>                j = i + 1
>             endif
>          else
>             j = j +1
>             if (j.gt.n) then
83c80
<                j = j - 1
---
>                j = i + 1
85c82
<          enddo
---
>          endif

3. 10/17/05

In add add-dHdSTm2.f and add-dHdSTm.f, dg is now rounded to two
decimal places and dg, dg and ds are written as f9.2, instead of as
f8.1. This is done to give better accuracy in computing ds and Tm.
In efn (efn.f) and nafold (misc.f), dg is written to two decimal
places as well.

4. 12/7/05

Error in creating local copy of sequence. No linefeed at the end of the
sequence file. mfold code (formid) is too stupid to read last line
without final line feed. Last line of reformatted sequence is missing.
Correction is to add a line feed at the end of the file.

In reformat-seq.sh, line feed is added in two locations.
106d105
<     echo '' >> ${FILE_PREFIX}-local.seq
124d122
< echo '' >> ${FILE_PREFIX}-local.seq