A practical tour of the whole bench — what a tool is for, the steps to run it on-device, and how to interpret what comes back, with the real thresholds the app uses.
Turn raw reads into contigs. When in doubt: Spine for long reads, the short-read assembler for Illumina.
The on-device overlap–layout–consensus assembler for Nanopore & PacBio HiFi — a complete circular chromosome and plasmids in ~90 s. Deep-dive →
Load long reads, choose Spine (top of the list), run. Optionally give an expected genome size; blank auto-estimates.
One circular contig near your expected size is ideal. Extra contigs are normal for plasmids or unresolved repeats — then check the trust card.
Short-read de-novo (multi-k de Bruijn, for Illumina); hybrid (long sets structure, short sharpens accuracy); and reference-guided, which maps to a reference and calls variants.
Pick the matching mode. Hybrid asks for both files; reference-guided asks you to choose a reference.
Short-read assemblies are inherently more fragmented. Reference-guided output is a variant table, not new contigs.
Scores trust on three axes — completeness (31 universal marker genes), contamination (per-contig composition), contiguity (N50, contigs, gaps) — as one green / amber / red verdict.
Opens automatically after assembly and lives under Analyze. Completeness needs annotation — run it first.
Green = near-complete, single-copy, single-organism. Amber = soft flag. Red = real problem (big gap or heavy duplication).
An experimental assembler that stores the genome as the weights of a small neural network. In active research & development and currently disabled — Spine is the assembler to use today.
Find the genes, then explore. Run annotation early — completeness, regulatory and the 16S call all need it.
Finds coding genes, tRNAs/rRNAs and other features and names them by function. Produces the annotation.gff3 many other tools depend on.
Analyze → Features, on an assembly.
A typical bacterium yields a few thousand CDS. "Hypothetical protein" = a gene found, no confident function.
A pan/zoom browser with tracks for GC, read depth, genes, variants, restriction sites and the raw sequence.
Jump by gene, locus or ACGT motif; drag to select to the base; switch linear/circular.
Depth dips flag weak spots; a sharp GC shift often marks a plasmid, island or prophage.
Open-reading-frame finder, codon-usage table, and a GC-skew plot. Run on any sequence or your assembly.
Analyze → ORF finder / Codon usage / GC skew.
GC skew usually flips sign at the replication origin and terminus — a check the chromosome is complete and oriented.
Predicts promoters (σ70 −10/−35), ribosome-binding sites, intrinsic terminators and operons.
Analyze → Regulatory; needs an assembly plus annotation.
Each call scores 0–100 — ≥75 strong, ≥50 moderate, below that a candidate.
Predicts where each protein ends up — cytoplasm, membrane, secreted, etc. Runs over the annotated proteins.
Analyze → Localization, on an annotated genome.
Surface-exposed and secreted proteins are the usual leads for vaccine/target work.
Maps DNA methylation (5mC / 6mA) and the methyltransferase motifs behind it (Dam GATC, Dcm CCWGG).
Import a Nanopore modBAM (MM/ML) or bedMethyl; set the min coverage to call a site (default 5×).
A motif near 100% methylated = an active methyltransferase; raise the coverage floor if calls look noisy.
The SNP/indel table from a reference-guided run, with the gene and amino-acid change for each call.
Opens from a reference-guided result; filter to protein-changing only. Export VCF.
Red = nonsense/frameshift, coral = missense, grey = synonymous. Watch low depth or mid allele-fraction calls.
Read-quality summary, k-mer spectrum, coverage plot and read pileup — the sanity checks before and after assembly.
Analyze → Read QC / Coverage on reads or an assembly.
A single clean k-mer peak = well-covered; a small second peak can mean a plasmid, repeat or contamination.
Everyday molecular-biology tools that work on any sequence — no genome required.
Multiple-sequence alignment plus a phylogenetic tree from the aligned sequences. Paste or import ≥2 sequences and run.
Analyze → Alignment.
Aligned columns show conserved vs variable sites; shorter branches / nearer tips = more similar.
A local similarity search — find where a query matches a subject (or your assembly).
Analyze → Sequence search / BLAST search.
High percent identity over most of the query length = a confident match.
Finds enzyme cut sites, simulates a digest, and draws the virtual gel. Set linear/circular, pick enzymes.
Analyze → Mol Bio → digest.
Single-cutters are the useful cloning sites; the gel shows fragment sizes as bands.
Designs Gibson & Golden Gate assemblies and site-directed mutagenesis, and draws the construct map.
Provide vector and insert (or the mutation).
Returns the junctions/overhangs and a finished circular product to export as FASTA.
Designs PCR primers and CRISPR guide RNAs, with an off-target scan. Pick a target region.
Analyze → Primer design / CRISPR.
Prefer matched melting temps with no strong dimers; for guides, a good on-target score and no close off-targets.
Scans protein domains, predicts secondary structure, and renders an interactive 3D model.
Analyze → Protein domains / Protein structure / PDB viewer.
Domain hits suggest function; the 3D structure is a model — a hypothesis, not a solved crystal.
Folds an RNA (or DNA-as-RNA) sequence into its likely secondary structure. Paste a sequence and run.
Analyze → RNA fold.
Lower (more negative) folding energy = more stable; stems are paired regions, loops are unpaired.
Two or more genomes side by side. Each needs at least two assembled genomes.
Aligns 2–8 genomes and shows synteny blocks, a BRIG-style identity ring, or a dotplot.
Analyze → Genome compare; pick ≥2 genomes.
Identity legend runs ≥98 / 90–98 / 75–90 / 50–75 / <50% / absent. Long colinear blocks = conserved backbone; breaks = rearrangements.
Builds a pairwise whole-genome SNP distance matrix, a tree (NJ/UPGMA), and single-linkage outbreak clusters at a cutoff you set.
Analyze → Phylogenetics / Epidemiology; select the runs.
Isolates under the cutoff cluster together — candidate transmission links. Tighter cutoff = more conservative.
Clusters genes across a set of genomes into core vs accessory. Needs ≥2 annotated genomes.
Analyze → Pangenome; set the protein-identity threshold.
Core = all, Soft-core ≥95%, Shell 15–95%, Cloud <15%. Big cloud = diverse population.
Measures selection on a gene (ω = dN/dS) and discovers over-represented motifs. dN/dS takes two orthologous CDS.
Analyze → Selection (dN/dS) / Motif discovery.
ω > 1.25 = positive selection, ≈1 = neutral, <0.85 = purifying, <0.5 = strong purifying.
What it is, what it carries, how it's typed — against bundled, versioned databases.
Research use only. Genotype predictions, not diagnostics — no substitute for laboratory susceptibility testing.
Identifies the organism from its 16S rRNA gene (~4,800-species database) and whole-genome ANI.
Analyze → Species ID. ANI needs the assembly; the 16S call needs annotation.
ANI ≥95% = same species (decisive). 16S: ≥98.7% species, 94.5–98.7% genus, 86.5–94.5% family. Disagree? ANI wins.
Screens for acquired resistance genes and known mutations, predicts affected drug classes, and flags high-stakes findings (carbapenem, colistin, fluoroquinolone). Database: AMRFinderPlus.
Analyze → Resistance genes, on an assembly.
Acquired gene called at ≥90% id & ≥60% coverage; mutations only on an exact catalogued change. Predicted, not measured.
Screens for virulence factors — toxins, adherence, secretion systems — grouped by category. Database: curated virulence set.
Analyze → Virulence, on an assembly.
A hit is called at ≥80% id & ≥70% coverage. Toxin hits are flagged.
Assigns a 7-locus sequence type (ST) and clonal complex. Schemes: K. pneumoniae, E. coli, S. aureus, P. aeruginosa.
Analyze → MLST, on an assembly.
ST assigned only on a 7/7 exact allele match; else NOVEL with the nearest ST — often a genuinely new type.
Identifies plasmid replicons (Inc groups) on your contigs. Database: PlasmidFinder.
Analyze → Plasmid typing, on an assembly.
A replicon is called at ≥90% id & ≥60% coverage. No replicon may just mean chromosomal or untypeable.
Maps the mobilome — prophages, IS/transposons, integrons, genomic islands — and flags resistance riding inside them. Needs annotation; contigs ≥20 kb.
Analyze → Mobile elements, on an assembly.
Each element carries high / medium / low confidence. Key alert: an AMR gene inside a mobile element can spread between strains.
Classifies reads or contigs by taxon and checks for a pure isolate vs a mixture.
Analyze → Metagenomics → Classify, on contigs or imported reads.
Pure isolate = one taxon ≥90%; Mixture = ≥2 taxa each >10% (contamination). Contaminants listed at ≥5%.
How data goes in and out, and how results stay reproducible and citable.
Read/write FASTA, FASTQ (incl. gzipped) and GenBank, and download references by accession. Assemblies, variant tables and reports all export to share off-device.
Reference databases carry a version manifest, and every typing/comparison result is stamped with the database name and version — reproducible and citable months later.
Process many isolates through a saved workflow, and generate publication-grade reports that carry their methods and citations with them.