GT-EVAL(1)

💖 GT-EVAL — the GenomeTools manual, bestie edition 💖

okay so you found gt-eval and you're about to compare your predictions against a reference and find out how well you ate. we love a self-aware moment. this page lives in the GenomeTools manual (that's the manual name, in case you were wondering 🎀), and yes, we're going to go through every single thing on it together. grab a snack. 🍓

---

✨ NAME ✨

gt-eval — compare annotation files and show accuracy measures (prediction vs. reference). it's giving "who did it better" but the answer is math. 💅

---

🌸 SYNOPSIS 🌸

gt eval reference_file prediction_file

two files, one vibe check. reference_file is the truth, prediction_file is what you're testing. go. 🦋

---

🫶 DESCRIPTION 🫶

hi bestie, here are all the flags. every one. no skips, no cuts, no cap:

• -nuc [yes|no] — evaluate at the nucleotide level. heads up though: memory consumption is proportional to the input file sizes, so if your files are huge your RAM is gonna be going through it 😭 (default: yes)
• -ltr [yes|no] — evaluate a LTR retrotransposon prediction instead of a gene prediction. note that all LTR_retrotransposon elements are considered to have an undetermined strand, that's just how it is (default: no)
• -ltrdelta [value] — set the allowed delta for LTR borders to be considered equal. how close is close enough? you decide (default: 20)
• -v [yes|no] — be verbose. tell me everything, i'm listening 👑 (default: no)
• -o [filename] — redirect output to specified file (default: undefined)
• -gzip [yes|no] — write gzip compressed output file (default: no)
• -bzip2 [yes|no] — write bzip2 compressed output file (default: no)
• -force [yes|no] — force writing to output file. she's pushy and we respect it (default: no)
• -help — display help and exit. the classic. 💀
• -version — display version information and exit

---

🌈 what it actually prints 🌈

the program shows sensitivity and specificity values for certain feature types — like gene, mRNA, and exon, you know, the usual suspects.

and for some feature types it also shows the number of missing and wrong features of that type. so what do those mean? let me spell it out, no gatekeeping:

• "missing" means the number of features of that type from the "reference" without overlap to a feature of that type from the "prediction". aka: it was there in the truth and your prediction said nothing. ghosted. 🙈
• "wrong" means the number of features of that type from the "prediction" without overlap to a feature of that type from the "reference". aka: you made it up. iconic of you, but wrong. 😭

---

🚨 REPORTING BUGS 🚨

found a bug? spill. report it to https://github.com/genometools/genometools/issues. don't suffer in silence, bestie, that's not the vibe. 💜

---

okay that's the whole page, you're fully caught up. this was GenomeTools 1.6.6, dated 11/09/2025, and that's the tea 🎀✨ go evaluate something and slay. 🫶