Genbank input file appears to have fewer records than expected.

[schellt]

Dear Augustus team, I wanted to retrain an Augustus model with following command using Augustus 3.3.3 (yes, not the latest version but as far as I could see this doesn't influence the issue): [I am shortening the paths for better readability]

autoAug.pl \
        --species=splicata \
        --genome=Splicata_final_assembly_smasked.fasta \
        --cdna=RNA_styela100.fasta \
        -v -v -v \
        --trainingset=splicata_round_1_2genome1_100.all.gff \
        --useexisting \
        --noninteractive

The error I receive is from optimize_augustus.pl. This is the last entry of STDOUT: 1 Running "perl optimize_augustus.pl --rounds=1 --species=splicata --trainOnlyUtr=1 --onlytrain=onlytrain.gb --metapars=splicata_metapars.utr.cfg train.gb --UTR=on > optimize.utr.out"... And this is the complete output of STDERR:

Genbank input file appears to have fewer records than expected.
This could be a consequence of using DOS (Windows) carriage return symbols at line breaks. at /cluster/software/augustus/augustus-3.3.3/scripts/optimize_augustus.pl line 462, <TRAINGB> chunk 33444.
failed to execute: perl optimize_augustus.pl --rounds=1 --species=splicata --trainOnlyUtr=1 --onlytrain=onlytrain.gb  --metapars=splicata_metapars.utr.cfg train.gb --UTR=on > optimize.utr.out!
failed to execute: perl autoAugTrain.pl -g=genome_clean.fa -s=splicata --utr -e=cdna.f.psl --aug=augustus.gff -w=autoAug -v -v -v --opt=1 --useexisting

I tried to track down the error myself but I ended up with more questions than answers.

If I understand correctly this error occurs, when $nloci < @namelines (line 461).

@namelines contains the lines from train.gb matching the pattern "LOCUS", as defined in line 408:

408     my @namelines = grep /^LOCUS   +/, @seqlist;

In lines 446-455 the file train.gb is processed and for every line $nloci is increased by one:

 446     while (<TRAINGB>) {
 447         my $gendaten = $_;
 448         m/^LOCUS +(\S+) .*/;
 449         my $genname = $1;
 450 
 451         $bucket = $bucketmap{$genname};
 452         my $handle = $fh[$bucket];
 453         print $handle $gendaten;
 454         $nloci++;
 455     }

In short: $nloci reflects the number of lines of train.gb and @namelines reflects the number of lines in train.gb, which match "LOCUS". Now I am wondering if this if statement in line 461 becomes true all the time, since @namelines is a subset of all lines of train.gb and thus never can be larger than the total number of lines of train.gb ($nloci).

Furthermore, train.gb was created while running autoAug.pl on a Linux machine and was never in contact with Windows.

Let me know, if you need any further information. Thank you very much in advance. Best, Tilman

[MarioStanke]

Please have a look at the input files to the last command, train.gb and onlytrain.gb manually. For example, they could be empty or have too few genes (LOCUS) for a successful training. If they contain no genes at all, a possible cause could be a mismatch between the gff and fasta file, e.g. different sequence names or the coordinates in the gff are from a different assembly version.

[schellt]

Hi @MarioStanke , thanks for the reply. I don't think that there are too few genes since I would expect to see the error message from line 413: Number of training sequences is too small.

$ grep -c "LOCUS" train.gb onlytrain.gb 
train.gb:151
onlytrain.gb:3431

A mismatch between gff and fasta would be impossible, since autoAug.pl was executed. Do you have any other idea? Do you have any comment on my thoughts on the code above?

[MarioStanke]

Another idea may be that your input training gene set contains too few UTRs. You could try to diagnose (eliminate this possibility) with --noutr.