cc-bioinfo · execution engine · v1.0.2
Not another AI chatbot.
A framework that actually does the work.
You do not install the environment, you do not write the code, and you do not sit and watch. Hand it the design and it builds its own conda environment, writes the analysis, fixes its own errors, produces publication-grade figures, and keeps going to an SCI manuscript draft. Going to clinic will not interrupt it.
What is bio-analyze?
bio-analyze is a structured knowledge framework: it encodes bioinformatics best practice as structured workflows with hard quality gates, guiding AI through an end-to-end analysis from raw-data quality control to a publication-ready SCI manuscript draft. It is the downstream execution engine of the cc-bioinfo platform, and one defining characteristic is that the framework itself contains no executable code — it is a 350 KB structured knowledge base, and the analysis code is generated on the spot by the AI and executed in your environment.
- 84,000+ lines of encoded domain knowledge across 34 knowledge files
- 6 standardized omics pipelines, plus knowledge-base coverage for ATAC-seq, WGCNA and Mendelian randomization
- 26 hours — the longest single uninterrupted run recorded in a real study
- 1,200+ automated tests passing, verified by double mutation testing with zero survivors
If you are a clinician or a graduate student, what does this mean?
The Steps and gates below sound engineering-heavy. From your side, here is what it actually looks like.
What you do not have to do
- No installing R, Python or Bioconductor — it builds the environment, bootstrapping conda from scratch if needed
- No writing code, no reading stack traces, no command line
- No sitting at the computer — it runs server-side and survives you closing the browser
- No memorising figure standards — 300 DPI, fonts and WCAG-compliant colourblind-safe palettes are checked against your target journal
What you must do
- Confirm the cell-type annotation — the step where AI is most likely to be wrong and domain knowledge most likely to save you
- Look at every figure — Step 4 stops and waits for your review; in real studies the most valuable defects were caught exactly this way
- Judge whether an odd result is biologically plausible — for example a gene that should not be highly expressed in your tissue
- Verify the manuscript — it gives you a draft; scientific accuracy and citations are yours to own
An honest sense of the timeline
Traditionally, from data in hand to a first draft takes most people months. On the platform, a comparative study with seven parallel evidence lines across six data layers completed in about 26 hours without interruption; a multi-dataset integration of 314,805 cells with spatial transcriptomics produced six main figures and a complete draft. But that does not mean you can submit two days later — you still need time to verify, to think, to revise. What it removes is the "getting things to run" part, not the "thinking it through" part.
How does the Step 0 to 7 workflow run?
| Step | What it does | Thinking tier |
|---|---|---|
| Step 0 | Project initialization: detect the omics type, create standard directories, load configuration | Standard |
| Step 1 | Dynamic analysis: multi-stage computation across phases 1 to N | Standard / Deep |
| Step 1.5 | Data quality audit: a five-dimension deep review; any critical finding triggers a hard stop | Ultrathink |
| Step 2 | Reflection: scoring how completely the scientific question was answered, translational assessment | Deep |
| Step 3 | Research summary: distilling core findings, planning figures | Ultrathink |
| Step 4 | Publication figures: SCI-standard output plus a visual review loop (this is where you review) | Deep |
| Step 5 | Technical report: a structured analysis record | Deep |
| Step 6 | Manuscript writing: 21 sub-steps plus five-step pre-submission cross-validation | Ultrathink |
| Step 7 | Submission preparation: DOCX, figure package, cover letter (user-triggered) | Standard |
Once you say "start the analysis" it executes immediately instead of asking "what next?"; fixable errors retry automatically up to three times; phases advance on completion. But the Step 1.5 quality audit and the Step 4 figure review stop and wait for you — those two gates cannot be skipped.
How is the four-tier thinking depth dispatched?
The governing idea: not every task deserves the same rigour. Renaming a file does not need a full chain of reasoning; annotating cell types does.
| Tier | Trigger | Behaviour |
|---|---|---|
| Quick | File operations, formatting | Immediate execution |
| Standard | Routine analysis steps | Standard reasoning |
| Deep | Parameter selection, method decisions, batch-effect assessment | Multi-option comparison |
| Ultrathink | Nine critical scenarios | Full chain: observe, hypothesize, validate, conclude |
The nine mandatory Ultrathink scenarios: cell-type annotation, the Step 1.5 quality review, differential expression interpretation, biological conclusion derivation, key manuscript sections, data validation, clustering-resolution selection, batch-effect assessment, and judging anomalous results.
Which omics types and pipelines are supported?
| Omics type | Platforms and tools | Standard pipeline |
|---|---|---|
| Single-cell RNA-seq | 10x Genomics, Seurat, Scanpy | QC, normalize, HVG, PCA, cluster, annotate, DEG |
| Spatial transcriptomics | Visium, MERFISH, Slide-seq, CODEX | QC, normalize, spatially variable genes, spatial clustering, deconvolution |
| Bulk RNA-seq | DESeq2, edgeR, limma | QC, align, count, normalize, DEG, enrichment |
| Proteomics | MaxQuant, DIA-NN, TMT | QC, normalization, imputation, differential proteins, pathways |
| Metabolomics | XCMS, MetaboAnalyst, MZmine | QC, normalization, PCA, differential analysis, enrichment |
| Lipidomics | Inherits metabolomics plus lipid-specific extensions | QC, normalization, PCA, differential analysis, lipid class |
Each pipeline defines mandatory steps and a required-figure checklist with live compliance checks: skipping doublet removal or batch correction raises a warning in flight, and missing mandatory figures trigger a hard stop before Step 4.
How are quality and data integrity guaranteed?
- The Step 1.5 five-dimension quality audit is inserted mandatorily between the end of computation and the start of biological interpretation: sample identity, batch effects, technical quality, biological plausibility and statistical assumptions. Any critical finding triggers a hard stop.
- Data-integrity iron rules forbid fabricated data, data leakage, silent substitution and raw data leaking into outputs. Generating simulated data is a critical error that halts the workflow outright.
- Executable validators rather than trusting the AI: workflow validation, state reconciliation and a 17-category consistency checker run every round and enforce their verdict through exit codes; the manuscript checker verifies every DOI against Crossref.
- Dual independent review: figures and manuscripts are adversarially re-checked by fresh independent sub-agents playing statistician, domain expert and sceptic — never self-approved.
- A self-learning experience store: 20 common error patterns ship with code-level fixes, and any fix that takes more than two rounds is saved automatically for reuse.
- An SCI figure knowledge base with a visual review loop: a 732-line standards file covering a chart-selection decision tree by sample size, eight bad-chart defences and 15+ reviewer checkpoints. The AI reads each rendered image, and your review gate must pass before the workflow advances.
What is bio-analyze not for?
| Scenario | Better alternative |
|---|---|
| Production pipelines processing hundreds of samples daily | Nextflow or Snakemake |
| Novel algorithm and methodology development | Direct programming |
| Pure upstream analysis such as assembly and alignment | BWA, GATK or SPAdes |
| Expecting it to replace your scientific judgment | Not applicable — key gates stop and wait for your call |
Stating the capability boundary honestly is part of the design philosophy: it does not replace statistical training, it does not guarantee the code runs correctly on the first attempt, and key checkpoints pause for a human decision.
How to get started
- Step 1: open a conversation on the platform. bio-analyze is pre-installed — you install nothing.
- Step 2: tell it your research question, where the data is and the target journal (or hand over the
TOPIC.ymlthat bio-design produced). - Step 3: say "start the analysis" and go do something else. It stops when it needs your call, and can reach you over IM.
/bio-analyze status --verbose # check progress
/bio-analyze continue # resume at sub-step granularity
/bio-analyze pipeline show # pipeline compliance status
/bio-analyze config set analysis.clustering_resolution 0.6 --global
/bio-analyze manuscript check --journal nature
/bio-analyze publish prep # submission packageThree-layer customization: project tier over user-global tier over framework tier, across 12 parameter domains and 50+ parameters. Save your preferences (a 15% mitochondrial threshold, a clustering resolution of 0.6) as global defaults and every new project inherits them — this is not a black box.
Frequently asked questions (FAQ)
How do cc-bioinfo, bio-design and bio-analyze relate to each other?
They are one platform plus two built-in engines, not three parallel products. cc-bioinfo is the platform and answers "where does it run"; bio-design is the design engine and answers "what study should I do"; bio-analyze is the execution engine and answers "how do I actually run it". In one sentence: cc-bioinfo is the stage, bio-design is the screenwriter, bio-analyze is the director and the whole cast.
Do I need to write R or Python code myself?
No. The AI generates, executes and debugs the code, and automatically creates a per-project conda environment, bootstrapping miniconda from scratch if none exists. What you do need is to understand your own scientific question and to make the key judgment calls: cell-type annotation confirmation, analysis direction, and figure review.
Do I have to sit and watch it run?
No. The analysis runs server-side and does not stop when you close the browser, so you can go to clinic or to theatre. When it needs your decision it can reach you through WeChat, DingTalk, Feishu or Telegram. The longest single uninterrupted run in a real case study was about 26 hours.
What happens if an analysis is interrupted halfway?
Run /bio-analyze continue. The system scans the completed outputs, diffs them against the plan, and resumes at exact sub-step granularity. Step 6 has 21 independent sub-steps that each write their own file, so only the failed one reruns.
Can I submit the generated manuscript directly?
Step 6 produces a draft; the researcher must verify scientific accuracy and complete the citations. The framework ships machine-readable compliance configurations for six major journals and automatically checks word limits, 300 DPI resolution, fonts and colourblind-safe palettes. Fabricating statistics or citations is strictly forbidden, and every cited DOI is checked against Crossref.
How do I try it, and how do I buy it?
Email moogtang@gmail.com to arrange a trial. Using it is straightforward: describe your analysis in plain language — no installation, no conda setup, no command line. The personal and enterprise editions are quoted on enquiry with no public price list — email moogtang@gmail.com; individual users in mainland China can also purchase directly through Xiaohongshu. See the purchase section.