Predict genes in a bacterial genome

What this involves

Bacterial gene prediction identifies coding sequences and translation initiation sites across assembled genomic contigs. Prodigal models coding potential, ribosomal binding site motifs, and translation frames directly from microbial nucleotide sequences without needing separate training sets.

Deriving functional insight from predicted genes requires translating coding sequences and screening them against reference protein sets. Doing this properly requires rigorous quality checks on the translated sequences, followed by sequence search with DIAMOND or MMseqs2 against databases such as UniProtKB, InterPro, or eggNOG to identify orthologues and functional domains.

The platform executes these procedures on its own research machine, where Prodigal, DIAMOND, MMseqs2, and SeqKit are already preinstalled on PATH. The research agent coordinates the commands, manages intermediate outputs in the work environment, and reads the results back without requiring local software installation.

How it goes

1. 1Validate assembly inputs with SeqKitInspect and summarise the uploaded nucleotide FASTA assembly using SeqKit 2.13.0 to verify sequence lengths and character alphabets.
2. 2Call coding genes using ProdigalExecute Prodigal 2.6.3 on the bacterial assembly to determine gene boundaries, generating coordinate records, nucleotide coding sequences, and translated protein FASTA files.
3. 3Clean and check translated proteinsStrip terminal stop codons from the translated proteins, verify reading frames by checking for internal stop codons, and evaluate residue ambiguity codes.
4. 4Search reference databases using DIAMOND or MMseqs2Run DIAMOND 2.1.12 or MMseqs2 to align predicted protein sequences against target databases, utilising multiple processor threads or background job management for long runs.
5. 5Assign orthology and functional annotationQuery UniProtKB, InterPro, or eggNOG to assign domain architectures, orthologous groups, and curated functional descriptions to matching alignments.

Start it here

Ask in plain language. The agent works on a Linux machine of its own — it installs and runs what the job needs, keeps every file, and reads the results back with the numbers explained. Nothing to set up, and a long run continues after you close the page.

What to watch for

• Reporting percent identity without query coverage and aligned length produces misleading orthologue calls.
• Comparing E-values across searches against different database sizes is invalid because E-values scale with database size whereas bit scores do not.
• Internal stop codons indicate a translation frame shift or pseudogene rather than a valid complete coding sequence.
• Deleting lowercase characters from masked assemblies shifts every downstream residue index instead of preserving coordinate alignment.
• Large reference databases required by DIAMOND can be tens of gigabytes and must be downloaded prior to running searches.

What it runs

Questions