A quick orientation to how the app actually works, what Omi can genuinely do across the computational life sciences, and where your data goes. For pricing and general FAQs, see the main site.
Five steps, no setup expertise required.
Grab the Windows installer from the download button above. It is unsigned by an independent publisher, so Windows SmartScreen may show a warning the first time — click More info → Run anyway to continue. This is expected and is explained again in the installer flow itself.
Create an account or sign in with email. Your account is what carries your subscription, saved lessons, and preferences across sessions.
On first login, Omi asks who you are (student / researcher / clinician / professional / curious) and what you mainly want to do (learn, analyze, build, or a mix). This calibrates how Omi teaches and paces itself — it is not a fixed quiz, and you can always ask Omi to go faster, slower, or skip the basics at any point.
The first time you use the app, it installs a local Python environment (Jupyter, the core scientific stack) so the embedded notebook can actually run code. This takes several minutes and needs an internet connection once — after that, it is ready every time you open the app. You will see a "Setting up your local lab" screen while this happens; it cannot be skipped, since nothing works without it.
Type what you want — "teach me what a p-value actually means," "analyze this FASTQ file I uploaded," or "build me a dashboard for this dataset." Omi will ask a clarifying question if it needs one, then get to work in the notebook, live, while it talks you through it.
How Omi decides what to run for a given task, with concrete examples of what's actually wired in today.
Omicsboard Lab does not ship one built-in "genomics module" or "statistics module" that tries to do everything itself. Instead, at its core is Omi — an AI that reasons about the specific task you have asked for, then reaches for whichever real, established, benchmark-standard tool actually does that job in that field, and runs it live, in a genuine Python notebook, on your own machine.
That means the honest answer to "does Omicsboard Lab support X?" is almost always: Omi will use whatever the field's own real tool for X is — not a proprietary stand-in. A few concrete, real examples of what is actually wired in today:
None of these are hardcoded once and forgotten, either — the notebook can install any Python package it needs on demand, the same way a person doing the analysis themselves would. Omi is not limited to a pre-approved list.
Three layers, and Omi is honest about which one you are in:
This is what "honest about what runs here vs. in the cloud" actually means in practice: Omi will say, in the moment, which of these three layers a given step falls into — not just at a marketing level, but as it works.
There is no mode to pick up front — just say what is wanted, and Omi adapts. Switch anytime, mid-conversation.
Omi teaches like a patient senior scientist — line by line, at your pace, on real data.
Try: "teach me what a p-value actually means, using a real dataset"
A real, publication-grade result on your own data — or on a topic Omi finds real data for.
Try: "analyze this FASTQ file I just uploaded"
Turn a method, paper, or idea into a real, working tool — not just a snippet.
Try: "build me a dashboard that compares these two conditions"
A technical inventory, not a highlight reel — see the main site for the short version.
The deepest-tested pillar today — these are named workflows, not the boundary of what's possible in this space. Ask Omi about anything genomics-adjacent, not only what's listed below.
Fetches sequences and genomes by accession directly from NCBI/ENA/SRA (verified against the real source, never guessed from memory). Local sequence handling, quality control, and trimming on real FASTQ data.
Finds genomes by real epidemiological context — country, host, or resistance profile — via NCBI Pathogen Detection (~100 curated surveillance species) and BV-BRC (bacteria and viruses, with a real country filter NCBI's own search doesn't support). Built for surveillance-style questions accession-only search can't answer.
Differential expression via PyDESeq2. Single-cell analysis via Scanpy — a gold-standard tool in its own right, not a substitute.
Taxonomy classification against the real NCBI 16S and SILVA reference databases (memory-safe, runs on the full dataset, not a toy subset). Diversity metrics, and an interactive Krona-style sunburst visualization.
Predicts the functional / metabolic pathway profile of a 16S community from its taxonomy — pulling the real KEGG gene content of each genus's own sequenced genomes, weighted by abundance. Reports its own coverage honestly (typically 40–60% of a real community), and is explicit that this is a prediction, not a direct measurement.
Local, targeted variant calling for gene/panel/chromosome-scale references — single-sample or across a whole cohort at once — no compiled aligner required, plus effect prediction and a real ClinVar lookup. Whole-genome, gold-standard variant calling routes to Galaxy.
Genome-wide copy-number profiling for shallow whole-genome sequencing — too low-coverage for SNP calling, but still real signal for gains and losses. Runs memory-safely one chromosome at a time, with every value measured from the reads, never estimated.
Screens a genome against the real, curated CARD (resistance) and VFDB (virulence) databases, and against the CGE PlasmidFinder database for plasmid replicon typing — with the database version always cited.
Real gene prediction and real homology-based ortholog clustering (sequence alignment, not composition heuristics), genome relatedness (ANI), and genome maps — across many genomes at once.
Assigns real functional domains (Pfam) and KEGG pathways (KO orthologs) to every gene family in a pan-genome, then tests statistically which functions are enriched in the core vs. accessory genome — not a guessed or hand-mapped annotation.
Real multiple sequence alignment and distance-based tree construction, with bootstrap support values genuinely computed. Selection analysis (dN/dS) on coding sequences.
General-purpose — the same underlying stack (pandas, numpy, scipy, statsmodels, scikit-learn) works on any tabular data you bring, not only genomics output.
Hypothesis testing, regression, correlation, and distribution fitting on whatever dataset you're working with — genomics or otherwise.
Cleaning, reshaping, and merging tabular data, and building publication-quality figures from it — the same visualization Omi uses throughout every other capability, available on its own.
Classification, regression, clustering, and dimensionality reduction on any feature table — general-purpose, not tied to a specific domain.
Writes and runs real multi-file software and interactive dashboards locally. Authors Dockerfiles, Nextflow/Snakemake, and CI configs for deploying at scale — actually deploying/running those is on the user's own infrastructure.
Searches the real literature (Europe PMC + OpenAlex) and pulls real public data (WHO, World Bank, NASA, GBIF, Ensembl, UniProt) — never invents a citation or a number.
Drafts a structured manuscript — sections, tables, and figures pulled from the actual analysis in the notebook — and checks every citation against CrossRef before it ships, rather than trusting a remembered reference.
Connect a free Galaxy account once. Omi submits R-only or heavy gold-standard workflows (e.g. DADA2 16S denoising) directly via the API — no manual clicking through Galaxy's website — and pulls the results back into the notebook.
Opens and reads real web pages inside the app — showing a reference page, guiding you through a site's own interface (like Galaxy's), or pulling in a page's content — without ever leaving the notebook for an external browser.
Files stay on the local machine. Everything runs in a Python notebook locally — uploaded files are read straight from disk by their local path. They are never uploaded to a server.
Only the conversation goes to the AI. Chat messages and the code Omi writes are sent through Omicsboard's own server-side proxy to the language model — no third-party API key ever lives inside the app itself, and raw data files are not part of that traffic.
Data Omi fetches (a public genome, a paper, a public dataset) comes from the real public source it names — NCBI, ENA, Europe PMC, Galaxy, and similar — not from a private copy held elsewhere.
"Setting up your local lab" keeps appearing — this only runs once and needs an internet connection; if it reappears after finishing once, it is a genuine environment check that found something missing and needs a moment to reinstall it. Let it finish rather than closing the app mid-way.
Something looks stuck or errored — from the app's title bar, open Help → Open Logs to see exactly what happened; this is the fastest way for support to help.
Need to talk to someone — from the title bar, Help → Contact / Get Help opens a form and a direct WhatsApp line, or reach out directly:
Looking for pricing details or general questions? Both live on the main site so there is one place to keep them current.