Use FragGeneScanRs as gene predictor instead of FragGeneScan++

[bmesuere]

FragGeneScan++ has been archived since 2021. FragGeneScanRs is maintained, faster, and ours.

Two commits: the swap, and the migration messages for people coming from the old setup.

The swap

The only thing in the way was -c, which the scripts passed and FragGeneScanRs does not accept. It only tunes FGS++'s chunked threading, so dropping it costs nothing. FragGeneScanRs also carries its own training data, so -r goes too.

That makes -f simpler: it now takes the binary, or a directory containing it, instead of a directory that had to hold both FGSpp and train. Without -f, setup looks for FragGeneScanRs on the PATH, which is where cargo, conda and the release tarballs all put it.

The setup check also got stricter. It now feeds a real read and requires a prediction to come back, rather than only checking that the binary starts. The old check piped in a single base:

printf '>a\nA' | ./FGSpp -s stdin -o stdout -w 0 -t illumina_10 -c 240

FragGeneScanRs panics on input that short (index out of bounds in viterbi.rs), worth reporting on that repo separately, but a check that only proves the binary starts was not worth much anyway.

Migration messages

Someone following the old instructions would otherwise hit a baffling failure: passing an FGS++ directory to -f makes setup look for FragGeneScanRs inside it, not find it, and report that invoking FragGeneScanRs failed.

• umgap-setup.sh recognises an FGS++ install, whether -f points at the directory or at the FGSpp binary, and says the predictor changed. A directory holding both binaries still picks FragGeneScanRs without complaining.
• umgap-analyse.sh does the same for an existing configuration directory. This is the likelier path: anyone who ran the old setup has an FGSpp symlink there and no reason to run setup again, so they would have got a bare "FragGeneScanRs not found" with no hint that anything changed.

Verification

Detection paths: -f as a directory, -f as a binary, found on PATH, and absent. Migration paths: -f old FGS++ directory, -f .../FGSpp, directory containing both binaries, config directory with an old FGSpp symlink, config directory with neither.

End to end on the bundled testdata:

umgap fastq2fasta testdata/A1.fq testdata/A2.fq \
  | FragGeneScanRs -s stdin -o stdout -w 0 -t illumina_10 -p 4 \
  | umgap prot2tryp | umgap filter

175 predictions from the 200 reads, tryptic peptides out the far end. Both scripts pass sh -n.

Note

Split out of #12 on request. That PR keeps the unrelated README corrections; the two touch disjoint parts of the README and merge cleanly in either order, which I verified locally.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.