This tutorial walks you through preparing your input data and interpreting the gene prediction output produced by Vipsania.
Vipsania works with any eukaryotic genomic FASTA file. For best results, follow these guidelines:
genome.fasta.gz) up to
500 MB are accepted, and the genome may
contain at most 300 Mbp.
Select the most specific clade that contains your organism:
| Dropdown label | Recommended for |
|---|---|
| Alveolata | Ciliates, apicomplexans, dinoflagellates (Plasmodium, Tetrahymena, …) |
| Amoebozoa | Amoebae and slime moulds (Dictyostelium, Entamoeba, …) |
| Arthropoda (other than insects) | Crustaceans, arachnids, myriapods; for insects use Insecta |
| Chlorophyta (green algae) | Green algae (Chlamydomonas, …) |
| Cnidaria | Corals, sea anemones, jellyfish (Nematostella, Hydra, …) |
| Discoba | Kinetoplastids, euglenids, heteroloboseans (Trypanosoma, Leishmania, Naegleria, …) |
| Echinodermata | Sea urchins, sea stars, sea cucumbers |
| Fungi | Fungi (Saccharomyces, Aspergillus, …) |
| Insecta | Insects (Drosophila, Apis, …) |
| Nematoda | Roundworms (Caenorhabditis, …) |
| Porifera | Sponges (Amphimedon, …) |
| Rhodophyta (red algae) | Red algae (Porphyra, Cyanidioschyzon, …) |
| Spiralia | Molluscs, annelids, flatworms |
| Stramenopiles | Diatoms, brown algae, oomycetes (Phaeodactylum, Phytophthora, …) |
| Streptophyta (land plants and relatives) | Land plants and charophyte algae (Arabidopsis, rice, …) |
| Tunicata | Tunicates (Ciona, …) |
| Vertebrata | Fish, amphibians, reptiles, birds, mammals |
| Other eukaryotes (none of the above) | Eukaryotes that belong to none of the clades listed above |
If your organism is not covered by any clade, choose Other eukaryotes. The model is applied as it is; this server does not finetune it on your genome (see Local installation and finetuning).
You will be redirected to a status page. Bookmark it; it updates automatically every 60 seconds. You will also receive an email when the job finishes or fails.
The status page also displays the MD5 checksum of your
uploaded genome file (computed from the raw compressed bytes).
Compare it with the output of
md5sum your_genome.fa.gz to confirm that the file was
uploaded intact.
After submission, your job enters a wait queue before execution begins. Wait time depends on current server load and can range from a few minutes to several hours. The status page shows whether your job is queued or running. Depending on resource availability, jobs may run on CPU or GPU hardware; GPU execution is substantially faster, so actual runtimes can vary significantly from the estimates below.
| Hardware | Genome size | Measured runtime |
|---|---|---|
| GPU (one A100) | 27–73 Mbp | 2–7 min |
| CPU (32 cores) | 27–73 Mbp | 1–7 h |
Runtime grows with genome size. Because any job may run on CPU, this server accepts genomes of up to 300 Mbp. Jobs time out after 72 hours.
Vipsania produces a gene annotation file in GFF3 format.
Each predicted gene has one transcript model (mRNA) with
annotated exon and CDS features.
Coding-sequence and protein-sequence FASTA files are also included.
The annotation was produced by the pretrained clade model
without finetuning; keep this in mind when you compare it
to published Vipsania accuracy figures, which are usually reported with
finetuning.
All result files are gzip-compressed (.gz). Decompress them
with gunzip before passing them to tools that do not accept
compressed input:
Many tools (genome browsers, BLAST, BUSCO, etc.) can read .gz
files directly without decompression — check your tool's documentation first.
Example GFF3 output (first few lines):
You can load this file directly into genome browsers such as JBrowse or IGV.
This web server never finetunes. We do not have the GPU capacity to run a training job for every submission, so there is no option for it. Finetuning adapts the model to the codon usage, repeat landscape and intron statistics of your species and may make the annotation somewhat more accurate. To get it, or for large genomes, repeated runs, or private data, run Vipsania locally (a GPU is strongly recommended):
or with the container image:
Without --finetune, the same commands reproduce what this
server does. See the
Vipsania GitHub repository
for full installation and usage documentation.