GitLab

Move html output folders into each networks output folder (i.e. networks_all,…

See merge request medema-group/BiG-SCAPE!9

Move html output folders into each networks output folder (i.e. networks_all, networks_all_lcs, ...) to prevent html output being overidden by multiple bigscape runs

Html visualization

See merge request medema-group/BiG-SCAPE!8

2.Change 'Network:' to 'Networks:'

Html visualization

See merge request medema-group/BiG-SCAPE!7

- Small speedup for LCS: trim pfam identifiers ("PF") when forming words for
difflib's SequenceMatcher
- Distance calculation stage now outputs number of gene where LCS seed starts
for both BGCs + correct orientation of the second BGC. This will be exported
eventually to the json file used for BiG-SCAPE's visualization

- Removed local and local-extended modes
- Implemented a new parameter called 'mode'. This can be set to 'global'
(default), 'lcs' and 'auto'
- New mode: 'auto'. Apply lcs mode only if any of the BGCs if fragmented (this
is known in the form of the 'contig_edge' annotation by AntiSMASH v4.0+)
- Code cleanup
- Removed MAFFT as an alignment option for domain sequences
- Removed BGCclassIndex column from network distance matrix

Bugfix: BGCs from Others class would get added twice to that class. This
happened once when finding BGCs with mixed-type annotations, then again if
any of its sub-annotations was also from the Others class.
Improvement: Only read fasta and pfd files once when calling GCFs (instead of
doing it at each cutoff)
Improvement: Pass reduced distance matrix to clusterJsonBatch (so we can get
rid of it quickly)
Improvement: Pass already sorted list of BGCs to clusterJsonBatch
Improvement: improved human readability of .js file
Improvement: Cleaned and re-ordered arguments (a bit)
Other: Changed clan-calling parameter to '--clans'

Note that they are currently different from the official bundle on MIBiG's page
as these ones have been further processed by antiSMASH. You can get this
version by downloading the gbk file in each entry's page

Lcs

See merge request !6

(cherry picked some changes. Will make further polishing)
# Conflicts:
#   Installation Guide.md
#   bigscape.py

Clans clustering

See merge request !5

1. Distance is average of average of pairwise distances between family members
2. Two arguments for clan cutoff: first is the cutoff value used for family assignment, second is the cutoff value for clans clustering
3. Fixed bug in distance calculation (max of similarities instead of famSimilarities)
4. Added clan_cutoff value to cutoffs if it's not there yet
5. Clans TSV file
6. Reworked clustering TSV file to take advantage of data structures

+ Some code cleanup

fixed typo and clarified installation instructions for pySAPC (probably will
have to work on that more in the future for people on Macs)

in their start/end positions.

Number of domains in each gene + gene orientation have to be obtained (after
having pfd information)

Prepare A_string and B_string.
Each is a list of concatenated pfam ids. Concatenation follows downstream orientation
of each gene

e.g.:
A_domlist = a b c d e f g   List of pfam ids as found in the BGC
dcg_a =   1  3  1  2        Number of domains per each gene in the BGC
go_a =    1 -1 -1  1        Orientation of each Gene
A_string = a dcb e fg       List of concatenated domains (note reverse order of bcd)

SequenceMatcher from difflib is used to find the largest common slice between
A_string and B_string. reverse(B_string) is also tested. Best orientation is kept

Extension of slices:
As expansion is relatively costly, a minimum of 3 overlapped genes are asked for
in the seed subcluster
 Extension occurs:
 For each upstream/downstream side:
  Find which is the BGC closest to the end.
  The slice of this BGC will be extended until the end
  The slice of the other BGC will be extended according to max_score
  (if both slices have the same length, the extension with the best score will
  be considered)
   Scoring:
   Input: slice of 'other' BGC and slice of extended BGC as reference.
   For each gene in 'other':
    Try to find gene in reference slice, starting in 'pos_y'
    If not found, decrease score by 'mismatch'
    If found, update score with 'match' + 'match_position'*'gap' and update
    'pos_y'
    If score >= max_score, update max_score and current position
   Output: position where max_score occurred, max_score.
   Match = +5
   Gap = -2
   Mismatch = -3

If smallest resulting slice has size 5 or bigger:
    If Biosynthetic Genes are in the final slices of both BGCs:
        Use slices for (normal) distance calculation.
    else
        Use the whole range of domains
else
    Use the whole range of domains

the one that extracted its sequences. Now all input files are opened only once.
Plus, this paves the way to having the Biosynthetic Genes annotated by
antiSMASH

Also, sanitize input cutoff values a bit more