An error occurred in convertalis when outputting alntmscore、fident and prob

[TigerWindWood]

Thank you for creating such an efficient tool.

Expected Behavior

I would like to obtain a foldseek result that includes alntmscore and the protein sequences of the hits

Current Behavior

$foldseek createdb query.pdb queryDB
$foldseek search queryDB targetDB aln_DB tmp
$foldseek convertalis queryDB targetDB aln_DB result.tsv --format-mode 4 --format-output query,target,alntmscore,fident,evalue,bits,prob,lddt,qstart,qend,tstart,tend,qcov,tcov,tseq
An error occurred: "Backtrace cigar is missing in the alignment result. Please recompute the alignment with the -a flag."
I followed the instructions and executed 'mmseqs align queryDB targetDB aln_DB alnNew -a', but the same error still occurred.
When I executed “$foldseek convertalis queryDB targetDB aln_DB result.tsv --format-mode 4 --format-output query,target,fident,evalue,bits,qstart,qend,tstart,tend,qcov,tcov,tseq,” it successfully output the results.
It seems that convertalis encountered an issue when outputting alntmscore, fident, and prob.

Foldssek Output (for bugs)

MMseqs Version: 9.427df8a
Substitution matrix aa:3di.out,nucl:3di.out
Alignment format 4
Format alignment output query,target,alntmscore,fident,evalue,bits,prob,lddt,qstart,qend,tstart,tend,qcov,tcov,tseq
Gap open cost aa:10,nucl:10
Gap extension cost aa:1,nucl:1
Database output false
Preload mode 0
Threads 150
Compressed 0
Verbosity 3
Exact TMscore 0

[=================================================================] 100.00% 1 eta -
Backtrace cigar is missing in the alignment result. Please recompute the alignment with the -a flag.
Command: mmseqs align queryDB targetDB aln_DB alnNew -a

Your Environment

foldseek Version: 9.427df8a

[permia]

hi, I come across the same problem.
As you said, It seems that convertalis encountered an issue when outputting alntmscore, fident, and prob.
Have you tried "mmseqs align queryDB targetDB aln_DB alnNew -a"

[martin-steinegger]

Could you please share an example to reproduce the issue?

[permia]

#!/bin/bash

# virus structure database
# /home/db/Nomburg_2023_structures/virus_3D/virus_3D
# /home/db/BFVD/bfvd
threads=12

if [ $# -ne 2 ] ; then
    echo "usage: foldseek_search.sh <queryDB> <targetDB>"
    exit 1
fi

queryDB=$(readlink -f "$1")
queryName=$(basename "$queryDB")

tmp=$(dirname "$queryDB")/tmp

targetDB=$(readlink -f "$2")
targetName=$(basename "$targetDB")

# the results in the query directory
result_dir=$(dirname "$queryDB")
resultDB="$result_dir"/"$queryName"_vs_"$targetName"

# search
foldseek search "$queryDB" "$targetDB" "$resultDB" "$tmp" \
-s 7.5 --threads "$threads" --alignment-type 2

# result conversion
foldseek convertalis "$queryDB" "$targetDB" "$resultDB" \
"$result_dir"/"$queryName"_vs_"$targetName"_tmp.tsv \
--threads "$threads" \
--format-mode 4 \
--format-output query,target,fident,alnlen,mismatch,gapopen,qstart,qend,tstart,tend,evalue,bits,qtmscore,ttmscore,alntmscore,rmsd,prob

# add header
awk -v OFS='\t' 'BEGIN {print "query", "target", "fident", "alnlen", "mismatch", "gapopen", "qstart", "qend", "tstart", "tend", "evalue", "bits"} {print}' \
 "$result_dir"/"$queryName"_vs_"$targetName"_tmp.tsv \
 > "$result_dir"/"$queryName"_vs_"$targetName".tsv

rm  "$result_dir"/"$queryName"_vs_"$targetName"_tmp.tsv 
rm -r "$tmp"

Here is my shell script. And the error messages:
Backtrace cigar is missing in the alignment result. Please recompute the alignment with the -a flag.
Command: mmseqs align queryDB targetDB resultDB alnNew -a.
So, should I rerun the mmseqs?

If I delete the following lines and using the default foldseek convertalis setting, It's OK.

--format-mode 4 \
--format-output query,target,fident,alnlen,mismatch,gapopen,qstart,qend,tstart,tend,evalue,bits,qtmscore,ttmscore,alntmscore,rmsd,prob

foldseek Version: 427df8a

[TigerWindWood]

hi, I come across the same problem. As you said, It seems that convertalis encountered an issue when outputting alntmscore, fident, and prob. Have you tried "mmseqs align queryDB targetDB aln_DB alnNew -a"

Yes, I have tried, but it still doesn't work.

[NatureGeorge]

A minimal example I have encountered is:

for pairwise comparison (ref #33)

foldseek createdb 1FZA.cif.gz 1FZG.cif.gz targetDB

foldseek easy-search targetDB targetDB fs_e_search.m8 tmp --alignment-type 1 --tmscore-threshold 0.0 --exhaustive-search 1 -e inf --format-mode 4 --chain-name-mode 1 --format-output query,target,fident,alnlen,qstart,qend,tstart,tend,qcov,tcov,alntmscore,qtmscore,ttmscore

If I remove exhaustive-search, then it goes fine.