# extrinsic information configuration file for AUGUSTUS
# include with --extrinsicCfgFile=filename
# date: 01.08.2006
# Mario Stanke (mario@soe.ucsc.edu)


# source of extrinsic information:
# M manual anchor (required)
# P protein database hit
# E est database hit
# C combined est/protein database hit
# D Dialign
# R retroposed genes
# T transMapped refSeqs

[SOURCES]
M RM

#
# individual_liability: Only unsatisfiable hints are disregarded. By default this flag is not set
# and the whole hint group is disregarded when one hint in it is unsatisfiable.
#
[SOURCE-PARAMETERS]
T individual_liability

#   feature        bonus         malus   gradelevelcolumns
#		r+/r-
#
# the gradelevel colums have the following format for each source
# sourcecharacter numscoreclasses boundary    ...  boundary    gradequot  ...  gradequot
# 

[GENERAL]
      start        1       1  M    1  1e+100  RM  1  1
       stop        1       1  M    1  1e+100  RM  1  1
        tss        1       1  M    1  1e+100  RM  1  1
        tts        1       1  M    1  1e+100  RM  1  1
        ass        1       1  M    1  1e+100  RM  1  1
        dss        1       1  M    1  1e+100  RM  1  1
   exonpart        1       1  M    1  1e+100  RM  1  1
       exon        1       1  M    1  1e+100  RM  1  1
 intronpart        1       1  M    1  1e+100  RM  1  1
     intron        1       1  M    1  1e+100  RM  1  1
    CDSpart        1       1  M    1  1e+100  RM  1  1
        CDS        1       1  M    1  1e+100  RM  1  1
    UTRpart        1       1  M    1  1e+100  RM  1  1
        UTR        1       1  M    1  1e+100  RM  1  1
     irpart        1       1  M    1  1e+100  RM  1  1
nonexonpart        1       1  M    1  1e+100  RM  1  1.15
  genicpart        1       1  M    1  1e+100  RM  1  1


#
# Explanation: 
# 
# The gff/gtf file containint the hints must contain somewhere in the last
# column an entry source=?, where ? is one of the source characters listed in
# the line after [SOURCES] above. You can use different sources when you have
# hints of different reliability of the same type, e.g. exon hints from ESTs
# and exon hints from evolutionary conservation information.
# 
# In the [GENERAL] section the entries second column specify a bonus for obeying
# a hint and the entry in the third column specify a malus (penalty) for
# predicting a feature that is not supported by any hint. The bonus and the
# malus is a factor that is multiplied to the posterior probability of gene
# structueres. 
# Example: 
#   CDS     1000  0.7  ....
# means that, when AUGUSTUS is searching for the most likely gene structure,
# every gene structure that has a CDS exactly as given in a hint gets
# a bonus factor of 1000. Also, for every CDS that is not supported the
# probability of the gene structure gets a malus of 0.7. Increase the bonus to
# make AUGUSTUS obey more hints, decrease the malus to make AUGUSTUS predict few
# features that are not supported by hints. The malus helps increasing
# specificity, e.g. when the exons predicted by AUGUSTUS are suspicious because
# there is no evidence from ESTs, mRNAs, protein databases, sequence
# conservation, transMapped expressed sequences.
# Setting the malus to 1.0 disables those penalties. Setting the bonus to 1.0
# disables the boni. 
# 
#       start: translation start (start codon), specifies an interval that contains
#              the start codon. The interval can be larger than 3bp, in which case
#              every potential start codon in the interval gets a bonus. The highest bonus is given
#              to potential start codons in the middle of the interval, the bonus fades off towards the ends.
#              For most species every start codon is an ATG, but if alternate translation_table's are used, then
#              all potential start codons are considered.	       
#        stop: translation end  (stop codon), see 'start'
#         tss: transcription start site, see 'start'
#         tts: transcription termination site, see 'start'
#         ass: acceptor (3') splice site, the last intron position
#         dss: donor (5') splice site, the first intron position
#    exonpart: part of an exon in the biological sense. The bonus applies only
#              to exons that contain the interval from the hint. Just
#              overlapping means no bonus at all. The malus applies to every
#              base of an exon. Therefore the malus for an exon is exponential
#              in the length of an exon: malus=exonpartmalus^length.
# 	     Therefore the malus should be close to 1, e.g. 0.99.
#        exon: exon in the biological sense. Only exons that exactly match the
#              hint get a bonus. Exception: The exons that contain the start
#              codon and stop codon. This malus applies to a complete exon
#              independent of its length.
#  intronpart: introns both between coding and non-coding exons. The bonus
#              applies to every intronic base in the interval of the hint.
#      intron: An intron gets the bonus if and only if it is exactly as in the hint.
#     CDSpart: part of the coding part of an exon. (CDS = coding sequence)
#         CDS: coding part of an exon with exact boundaries. For internal exons
#              of a multi exon gene this is identical to the biological
#              boundaries of the exon. For the first and the last coding exon
#              the boundaries are the boundaries of the coding sequence (start, stop).
#         UTR: exact boundaries of a UTR exon or the untranslated part of a
#              partially coding exon.
#     UTRpart: The hint interval must be included in the UTR part of an exon.
#      irpart: The bonus applies to every base of the intergenic region. If UTR
#              prediction is turned on (--UTR=on) then UTR is considered
#              genic. If you choose against the usual meaning the bonus of
#              irparts to be much smaller than 1 in the configuration file you
#              can force AUGUSTUS to not predict an intergenic region in the
#              specified interval. This is useful if you want to tell AUGUSTUS
#              that two distant exons belong to the same gene, when AUGUSTUS
#              tends to split that gene into smaller genes.
# nonexonpart: intergenic region or intron. The bonus applies to very non-exon
#              base that overlaps with the interval from the hint. It is
#              geometric in the length of that overlap, so choose it close to
#              1.0. This is useful as a weak kind of masking, e.g. when it is
#              unlikely that a retroposed gene contains a coding region but you
#              do not want to completely forbid exons.
#   genicpart: everything that is not intergenic region, i.e. intron or exon or UTR if
#              applicable. The bonus applies to every genic base that overlaps with the
#              interval from the hint. This can be used in particular to make Augustus
#              predict one gene between positions a and b if a and b are experimentally
#              confirmed to be part of the same gene, e.g. through ESTs from the same clone.
#              alias: nonirpart
#
# Any hints of types dss, intron, exon, CDS, UTR that (implicitly) suggest a donor splice
# site allow AUGUSTUS to predict a donor splice site that has a GC instead of the much more common GT.
# AUGUSTUS does not predict a GC donor splice site unless there is a hint for one.
# 
# Starting in column number 4 you can tell AUGUSTUS how to modify the bonus 
# depending on the source of the hint and the score of the hint. 
# The score of the hints is specified in the 6th column of the hint gff/gtf.
# If the score is used at all, the score is not used directly through some
# conversion formula but by distinguishing different classes of scores, e.g. low
# score, medium score, high score. The format is the following:
# First, you specify the source character, then the number of classes (say n), then you
# specify the score boundaries that separate the classes (n-1 thresholds) and then you specify
# for each score class the multiplicative modifier to the bonus (n factors). 
# 
# Examples:
# 
# M 1 1e+100
# means for the manual hint there is only one score class, the bonus for this
# type of hint is multiplied by 10^100. This practically forces AUGUSTUS to obey
# all manual hints.
# 
# T    2       1.5 1 5e29
# For the transMap hints distinguish 2 classes. Those with a score below 1.5 and
# with a score above 1.5. The bonus if the lower score hints is unchanged and
# the bonus of the higher score hints is multiplied by 5x10^29.
# 
# D    8     1.5  2.5  3.5  4.5  5.5  6.5  7.5  0.58  0.4  0.2  2.9  0.87  0.44 0.31  7.3
# Use 8 score classes for the DIALIGN hints. DIALIGN hints give a score, a strand and
# reading frame information for CDSpart hints. The strand and reading frame are often correct but not
# often enough to rely on them. To account for that I generated hints for all
# 6 combinations of a strand and reading frame and then used 2x2x2=8 different
# score classes:
# {low score, high score} x {DIALIGN strand, opposite strand} x {DIALIGN reading frame, other reading frame}
# This example shows that scores don't have to be monotonous. A higher score
# does not have to mean a higher bonus. They are merely a way of classifying the
# hints into categories as you wish. In particular, you could get the effect of
# having different sources by having just hints of one source and then distinguishing
# more scores classes.
# 
# 
# Future plans:
# - Add fuzzy intron hints. Introns get a bonus only when they approximately
# have the same boundaries as in the hint.
# - Make the splice site hints fuzzy also. Allow a hint interval that contains a
# likely splice site, as opposed to only an individual position.
# - Write a program that automatically optimizes the boni and mali given an
# annotated test set of genes and hints for that set of sequences.