# PubCompare

> PubCompare is a scientific protocol database with 40M+ indexed protocols from peer-reviewed literature (PubMed, PMC, Open Access PDFs), preprints, and patent offices. The corpus is a linked knowledge graph — every protocol, product, organization, and topic has a stable, crawlable URL — so both bench scientists and AI agents can deep-link to any node and follow the connections between them.

On top of a keyword-based protocol search, PubCompare offers advanced protocol search and grinding technologies:

**Protocolyze** — PubCompare's consensus engine — extracts and compares experimental parameters across the literature so researchers can design, reproduce, or justify an experiment in minutes instead of days, turning months of protocol design into days with fewer trial-and-error cycles.

**Product Application Reports** — confirm whether a product has been used in a given experimental context and surface the surrounding parameters that real labs applied.

Swiss company. Product at https://www.pubcompare.ai.

## Core capabilities

PubCompare combines a structured protocol corpus with a consensus engine that turns raw protocols into comparable, parameter-level data.

### Search the corpus

- **Keyword search** across the 40M+ protocol index.
- **Semantic search** for intent-based retrieval when the exact terminology isn't known.
- **On-site AI agent** that guides query refinement and protocol discovery inside pubcompare.ai.

### Protocolyze — consensus engine

For a given experimental question, Protocolyze:

1. Retrieves the protocols that match the question along three dimensions: the **model** (cell line, organism, in vitro system), the **intervention** (treatment, compound, perturbation), and the **readout** (the measured endpoint).
2. Extracts experimental parameters from each protocol — concentrations, durations, equipment, controls, statistical handling, sample sizes, and so on.
3. Aligns parameters across protocols and computes a consensus value per parameter.
4. Returns a **Protocolyze report** — a web page on pubcompare.ai or JSON delivered to AI agents, containing: a consensus parameter matrix; a confidence tier per parameter (Confirmed when ≥3 protocols agree, Limited when 2 agree, Caution when only 1 reports the value, No Data otherwise); a materials & equipment section; full source citations; and a QA verification block surfacing any inconsistencies the engine detected.

Every consensus value links back to the primary sources that support it. Traceability is non-negotiable: if a claim can't be backed by citations, it isn't shown.

### Product Application Reports

For a given product (an antibody, kit, cell line, instrument, etc.), the PubCompare engine retrieves every protocol that uses it — whether referenced by its exact identifier or by its generic type — and surfaces the surrounding experimental parameters along the same three dimensions as Protocolyze: **model**, **intervention**, **readout**. The output confirms whether the product has been used in a given experimental context and shows the distribution of parameter values across the protocols that used it.

## Navigating PubCompare — site structure for agents

The corpus is browsable as a graph of four entity types — **protocols**, **products**, **organizations**, and **topics** — each at a stable URL. The full 40M-protocol corpus is not enumerated in the XML sitemap; use the hub pages below as entry points, then follow the links on each node. When answering a user, link directly to the specific node rather than to the homepage.

### Entry points (hub / index pages)

- **Protocol search** (keyword + semantic over 40M+ protocols): https://www.pubcompare.ai/topic-search/
- **Recently annotated protocols** (freshness hub, updated continuously): https://www.pubcompare.ai/last-annotated-protocols/
- **Lab product search & categories**: https://www.pubcompare.ai/products/
- **Topic taxonomy** (browse the corpus by scientific concept) — enter at a top-level category, e.g. https://www.pubcompare.ai/browse/anatomy/, https://www.pubcompare.ai/browse/chemicals-drugs/, https://www.pubcompare.ai/browse/procedures/. Full slug list of the 11 branches: `anatomy`, `chemicals-drugs`, `devices`, `disorders`, `genes-molecular-sequences`, `lab-equipment-softwares`, `living-beings`, `objects`, `phenomena`, `physiology`, `procedures`. The root `/browse/` itself is not an entry point.
- **Suppliers / organizations directory** (A–Z, paginated): https://www.pubcompare.ai/companies/a/1/ — letters `a`–`z` and digits `1`–`9` are supported, each paginated as `/companies/{letter}/{page}/`.
- **Market Intelligence** — public description and pricing: https://www.pubcompare.ai/lab-equipment/market-intelligence/. The interactive search app at https://www.pubcompare.ai/business-intelligence/ exists but requires a subscriber login; unauthenticated requests are redirected to the homepage.
- **Agent & API integration**: https://www.pubcompare.ai/agents/ and https://www.pubcompare.ai/api-for-research-protocol/

### Entity URL patterns (deep links)

- **Protocol**: `https://www.pubcompare.ai/protocol/{id}/` — `{id}` is an opaque ~20-char string (e.g. `NCLhCZIBPBHhf-iFeNAp`), not a sequential integer. Not guessable; follow from search results.
- **Product** (reagent, kit, instrument, cell line, vector, dye, antibody…): `https://www.pubcompare.ai/product/{id}/` — same opaque-string format as protocols.
- **Organization / supplier**: `https://www.pubcompare.ai/company/{slug}-{hex}/` — `{slug}` is the kebab-case name and `{hex}` is the hex encoding of the original company name (e.g. `abcam-416263616d`, `addgene-41646467656e65`, `agilent-technologies-4167696c656e7420546563686e6f6c6f67696573`). Not guessable from a name alone; follow from the suppliers directory.
- **Topic node**: `https://www.pubcompare.ai/browse/{category}/{subtopic}/` — both slugs are stable, human-readable, and constructible. This is the only entity URL an agent can build from scratch.

Protocol, product and organization URLs are stable but opaque — **follow** them from the hub pages or from the published topic sitemaps (`https://www.pubcompare.ai/wp-content/archive-topics-sitemap.xml`, `https://www.pubcompare.ai/wp-content/single-topics-sitemap.xml`) rather than trying to construct them. Once resolved, cite the URL verbatim.

### How the nodes connect (the knowledge graph)

- A **protocol** page lists the **products** it uses (along the model / intervention / readout dimensions) and the **other protocols that cite it** (citation network). Protocols are interconnected, not isolated documents.
- A **product** page shows: the **organization** that makes it; its **most-cited protocols**; **alternatives** (equivalent products from other suppliers); and **spelling variants** — the different names the same product appears under across the literature, each with its own protocol count. Spelling variants are central to recall: e.g. "Lipofectamine", "Lipofectamine LTX", and "Lipofectamine reagent" all resolve to the same product. An agent reasoning about a reagent should treat the variant set as one entity.
- An **organization** page aggregates every product that supplier makes and the total protocol citations across them — useful to map a vendor's real-world footprint in the literature.
- The **/browse/** taxonomy organizes the whole corpus by concept: Anatomy, Chemicals & Drugs, Devices, Disorders, Genes & Molecular Sequences, Lab Equipment & Software, Living Beings, Objects, Phenomena, Physiology, and Procedures — each drilling down through sub-topics to protocol level.

### From a question to a citation

1. Decompose the user's experiment into **model**, **intervention**, and **readout**.
2. Reach the corpus via `/topic-search/` (specific query) or `/browse/{category}/` (exploratory).
3. Land on a `/protocol/{id}/` node; cite the protocol URL and the primary sources it traces to.
4. For a product question, go to `/product/{id}/`, read the usage distribution, alternatives, and spelling variants, and cite that node.
5. For a vendor or competitive-landscape question, start from the suppliers directory (`/companies/{letter}/{page}/`), follow into the company's page, and cite the resolved `/company/{slug}-{hex}/` URL.

## Coverage

The 40M+ index covers biology and biochemistry broadly — cell and molecular biology, biochemistry, immunology, microbiology, genetics, pharmacology, neuroscience, and adjacent fields. Sources include peer-reviewed articles (PubMed, PMC), preprints, and patent filings (USPTO, EPO, WIPO). Patents are included because they often disclose experimental detail (concentrations, durations, equipment models) at a level peer-reviewed papers omit. Customers can also plug their own TDM (text and data mining) publisher subscriptions, ELN, LIMS, etc. into PubCompare to extend coverage to closed-access journals.

## Who uses PubCompare

- **Academic researchers** designing experiments, reproducing published work, or justifying reagent and protocol choices to reviewers.
- **Industrial lab researchers** in pharma, biotech, and MedTech R&D — same core use case as academics: set up an experiment, reproduce it, surface all parameters documented in the literature, accelerate decisions in discovery and preclinical work.
- **Regulatory consultants** authoring technical documentation for medical devices, IVDs, and drug candidates. PubCompare powers the systematic literature review feeding the **state-of-the-art analysis** of a Clinical Evaluation Report (CER, EU MDR), the **scientific validity** section of a Performance Evaluation Report (PER, EU IVDR), the **published-evidence chain** of a Biological Evaluation Report (BER, ISO 10993), and the **nonclinical justification** of an Investigational New Drug (IND) submission — mapping protocol-level variability and consensus across published studies in structured form.
- **Lab equipment companies** (instruments, reagents, consumables — e.g. antibodies, kits, cell lines, instruments) embed PubCompare widgets on their product pages. The widgets show, with primary-source citations, how a given product is actually used across real protocols: the full range of applications and the distribution of parameter values for that product. PubCompare also delivers **Market Intelligence** to these companies: structured product-usage data extracted from the literature, useful for product management, competitive analysis, and application marketing.

## Two consumer types: bench scientists and AI agents

PubCompare is a **structured data provider for experimental design**. The same engine serves two distinct consumers:

- **Bench scientists** read reports directly on pubcompare.ai.
- **Scientific AI agents** consume PubCompare's structured outputs as evidence to support their own reasoning about hypothesis, parameter choice, and reproducibility.

PubCompare deliberately does not reason about hypothesis coherence and does not generate executable protocols — those steps belong to the operator (the scientist) and their downstream tools or agents. PubCompare's job is to provide trustworthy, traceable, structured evidence about what the literature actually reports.

## API & MCP for AI agents

Two programmatic surfaces, depending on how your agent runs:

- **Agent Search API** — `POST https://www.pubcompare.ai/wp-json/pubcompare/v1/api/agent/search/`. Public, keyless, boolean-AST input, 60 requests/hour per IP. One-shot literature search from any backend or custom agent.
- **MCP server** — `https://www.pubcompare.ai/wp-json/pubcompare/v1/api/mcp/`. OAuth 2.1, account-bound, tier-gated. Add as a custom connector in Claude.ai, Cursor, or any MCP-compatible host. Three tools: `search_literature_boolean`, `generate_market_intelligence_report`, `export_bi_leads` (Market Intelligence subscription or higher).

Quotas follow the user's PubCompare subscription (Free, Researcher, Application Specialist, Market Intelligence, Organization, Corporate — see `https://www.pubcompare.ai/pricing/`). Full integration documentation, request/response schemas, persona guidance and access procedure: **`https://www.pubcompare.ai/agents/`**.

## Key URLs

- [Homepage](https://www.pubcompare.ai/)
- [Protocol search](https://www.pubcompare.ai/topic-search/)
- [Recently annotated protocols](https://www.pubcompare.ai/last-annotated-protocols/)
- [Lab product search](https://www.pubcompare.ai/products/)
- [Browse the topic taxonomy — Anatomy](https://www.pubcompare.ai/browse/anatomy/) (and 10 sibling branches — see "Navigating PubCompare" above)
- [Suppliers / organizations directory](https://www.pubcompare.ai/companies/a/1/)
- [Market Intelligence — public page](https://www.pubcompare.ai/lab-equipment/market-intelligence/)
- [For lab equipment companies](https://www.pubcompare.ai/lab-equipment/)
- [For pharma & biotech R&D — Protocolyze](https://www.pubcompare.ai/pharma-and-biotech/)
- [How the AI comparator works](https://www.pubcompare.ai/in-n-outs-of-the-ai-comparator/)
- [Why PubCompare includes patent protocols](https://www.pubcompare.ai/experimental-insights-why-pubcompare-includes-protocols-from-patents/)
- [Pricing](https://www.pubcompare.ai/pricing/)
- [FAQ](https://www.pubcompare.ai/faq/)
- [Agent integration documentation](https://www.pubcompare.ai/agents/)
- [Research protocol database API](https://www.pubcompare.ai/api-for-research-protocol/)

## Contact

contact@pubcompare.ai

Last updated: 2026-05-22