Methodology
Data Management & Methodology
How VivoSeq collects, validates, and synthesizes biopharma intelligence from authoritative public sources.
46,204
Active Trials
17,558
Diseases
24,964
Programs
4,255
Companies
260
Publications
4,363
FDA Approvals
2,074
Branded Drugs
Sources
Authoritative Data Sources
VivoSeq ingests data from eight authoritative sources, each selected for its reliability and coverage of the biopharma landscape.
ClinicalTrials.gov
PrimaryClinical trial registrations, status updates, enrollment data, phase transitions, sponsor information
SEC EDGAR
PrimaryCompany financial filings (10-K, 10-Q, 8-K), revenue, employee counts
PubMed / MEDLINE
PrimaryBiomedical literature, clinical study publications, systematic reviews
openFDA
PrimaryDrug approval dates, labeling, adverse events, manufacturer information
FDA Orange Book
PrimaryApproved drug products (NDAs), brand/generic linkage, patent data, manufacturer identification
FDA Purple Book
PrimaryLicensed biological products (BLAs), biosimilar reference products, manufacturer identification
USPTO PatentsView
PrimaryPatent grants and expiration estimates for approved drugs
Company profiles (FMP, Wikidata)
SupplementaryMarket cap, CEO, headquarters, founding year, geographic coordinates
ClinicalTrials.gov
PrimaryClinical trial registrations, status updates, enrollment data, phase transitions, sponsor information
SEC EDGAR
PrimaryCompany financial filings (10-K, 10-Q, 8-K), revenue, employee counts
PubMed / MEDLINE
PrimaryBiomedical literature, clinical study publications, systematic reviews
openFDA
PrimaryDrug approval dates, labeling, adverse events, manufacturer information
FDA Orange Book
PrimaryApproved drug products (NDAs), brand/generic linkage, patent data, manufacturer identification
FDA Purple Book
PrimaryLicensed biological products (BLAs), biosimilar reference products, manufacturer identification
USPTO PatentsView
PrimaryPatent grants and expiration estimates for approved drugs
Company profiles (FMP, Wikidata)
SupplementaryMarket cap, CEO, headquarters, founding year, geographic coordinates
Safeguards
Data Quality Safeguards
The structural protections every data point passes through before it reaches the database: a source-authority hierarchy, validation rules, entity resolution, and guarded AI enrichment. Higher-authority sources win, and enrichment never overwrites a verified value.
FDA Orange Book, Purple Book, openFDA
Highest authority. Overrides all other sources for drug approvals, manufacturer identity, and brand/generic linkage.
ClinicalTrials.gov
Authoritative for trial registrations, phases, enrollment, and sponsor relationships. Not authoritative for commercial drug information.
SEC EDGAR
Authoritative for US-listed public company financials, revenue, and employee counts.
PubMed / MEDLINE
Publication linkage and scientific context. Supplements but does not override regulatory or trial data.
Wikidata, Financial Modeling Prep
Company metadata and market data. Lowest-priority structured sources.
AI Enrichment
Used only when no authoritative source exists. Tagged separately. Requires high-confidence threshold.
Brand–Generic Consistency
Brand names must be meaningfully different from generic names. Prevents raw ingredient names from being stored as brand names.
Drug–Company Verification
Drug-company associations are validated against known pharmaceutical manufacturers. Prevents trial sponsors like hospitals from being listed as drug manufacturers.
Phase Normalization
Phase values from different sources (PHASE3, Phase 3, phase III) are mapped to a canonical format: Phase 1, Phase 2, Phase 3, Approved.
Status Standardization
Trial statuses are standardized from raw API formats to display-ready values for consistent filtering and display.
Comparator Arm Filtering
Clinical trial comparator arms (placebo, standard of care, investigator's choice) are identified and excluded from pipeline drug listings.
Ingestion Safeguards
New data enriches but never overwrites. If a field already has a verified value, incoming data from a lower-authority source cannot replace it.
Company Consolidation
- Canonical name mapping for 40+ major pharma companies and all known subsidiaries
- Corporate suffix normalization (Inc., LLC, Corp., Ltd., Pharmaceuticals, Therapeutics)
- Automatic subsidiary detection and parent company resolution
- Duplicate ticker detection across exchanges
Drug Deduplication
- Normalization of dosage forms, salt names, and formulation variants
- Brand-to-generic linkage via FDA Orange Book data
- Combination therapy identification and separation from standalone drugs
- Dose-variant slug consolidation (e.g., drug-10mg, drug-20mg)
Disease Mapping
- MeSH term normalization for standardized disease naming
- Cross-reference between ClinicalTrials.gov condition names and standardized nomenclature
- Plural/singular variant merging and case normalization
- Junk indication filtering (headache, sedation, complication excluded from landscapes)
Gap Detection
Identifies investigational compounds where authoritative FDA sources don't provide complete manufacturer, brand name, or modality information.
AI Query
Queries language models with pharmaceutical domain knowledge to identify the correct manufacturer, brand name, and drug modality.
Cross-Reference
AI responses are cross-referenced against existing database records to check for consistency and prevent duplicate entries.
Confidence Filter
Only high-confidence answers are applied. Low-confidence results are discarded rather than introducing noise into the dataset.
Provenance Tag
All AI-enriched data is tagged separately from FDA-verified data for full transparency. Users can distinguish verified from enriched records.
Automated Review
Research Agents
VivoSeq runs a set of automated research agents — the digital equivalent of a data-curation desk. Each runs on its own schedule, reviews a slice of the database, and either corrects issues directly (with provenance locks) or flags them to an internal review queue. Select an agent to see its purpose, what it checks, and how often it runs.
QA Sweep
The continuous data-quality agent. It re-checks records against authoritative sources and corrects issues at the root — automatically when a fix is unambiguous, and through a human review queue when it needs judgment. Every automatic fix is recorded with a provenance lock, so a later data re-import can't silently revert it.
Cadence
Daily
Scope
Drugs, companies & calculated metrics
On finding
Auto-fixes safe cases; routes the rest to review
Checks performed
4Spot Check
Emulates a human QA reviewer doing daily spot-checks across companies, diseases, drugs, and trial links — the kind of obvious errors a person would catch at a glance. Discrepancies are surfaced to the review queue rather than changed automatically.
Cadence
Daily
Scope
Companies, diseases, drugs & trial links
On finding
Flags discrepancies to the review queue
Checks performed
5Page Audit
An LLM “eyeball” that loads live pages the way a visitor would and flags anything that looks broken or nonsensical before a user sees it. It samples each page type, reads the rendered content, and reports obvious problems for review.
Cadence
Weekly
Scope
Live rendered pages (sampled per type)
On finding
Reports broken or nonsensical content for review
Checks performed
4Deduplication Scan
Finds duplicate entities across companies, drugs, and diseases using fuzzy name matching, ticker comparison, and brand–generic cross-referencing, so the same real-world entity isn't counted twice.
Cadence
Weekly
Scope
Companies, drugs & diseases
On finding
Queues duplicate clusters for merge review
Checks performed
4Phase Validation
Cross-references pipeline phase assignments against FDA approval records and clinical-trial phase data to catch misclassified drugs (e.g. an approved therapy still shown as Phase 1).
Cadence
Weekly
Scope
Pipeline phases vs FDA & trial records
On finding
Flags misclassified phases for review
Checks performed
4Content Audit
The “is this obviously stupid?” agent. Rather than checking named patterns, it reads what users actually see — each disease and the top drugs we rank as its pipeline — and an LLM flags anything a knowledgeable analyst would immediately call wrong. It covers the whole pipeline surface, not a sample, and routes flags to the review queue (never auto-deletes).
Cadence
Weekly
Scope
Every meaningful disease's ranked pipeline
On finding
Flags implausible drug–disease pairs to review
Checks performed
4Intelligence
Calculated Metrics
Proprietary intelligence metrics derived from raw public data. All scored on a 0–100 scale.
Competitive Density
DiseaseHow crowded a disease landscape is — combining active programs and clinical trials into a single activity score
Pipeline Velocity
CompanyThe rate of new trial activity — how many trials have started recently relative to historical norms
Mechanism Concentration
DiseaseWhether programs in a disease area are converging on the same approach or exploring diverse mechanisms
Trial Completion Rate
CompanyThe proportion of trials that reach completion versus total terminal outcomes — a signal of execution quality
Attrition Rate
DiseaseThe rate at which programs are terminated or withdrawn — a risk signal for the disease area
R&D Momentum
CompanyA composite score reflecting a company’s recent trial starts, active programs, and pipeline breadth
Pipeline Concentration
CompanyHow diversified a company’s pipeline is across therapeutic areas — measured using the Herfindahl-Hirschman Index
Pipeline Freshness
CompanyThe average age of trials in a company’s portfolio — younger pipelines suggest active investment
Momentum Score
Therapeutic AreaA therapeutic area or disease-level signal combining recent trial starts, company participation, and overall activity
Coverage Gap
DiseaseThe relationship between disease prevalence and pipeline activity — highlighting potential unmet need
Pipeline Velocity
CompanyThe rate of new trial activity — how many trials have started recently relative to historical norms
Trial Completion Rate
CompanyThe proportion of trials that reach completion versus total terminal outcomes — a signal of execution quality
R&D Momentum
CompanyA composite score reflecting a company’s recent trial starts, active programs, and pipeline breadth
Pipeline Concentration
CompanyHow diversified a company’s pipeline is across therapeutic areas — measured using the Herfindahl-Hirschman Index
Pipeline Freshness
CompanyThe average age of trials in a company’s portfolio — younger pipelines suggest active investment
Competitive Density
DiseaseHow crowded a disease landscape is — combining active programs and clinical trials into a single activity score
Mechanism Concentration
DiseaseWhether programs in a disease area are converging on the same approach or exploring diverse mechanisms
Attrition Rate
DiseaseThe rate at which programs are terminated or withdrawn — a risk signal for the disease area
Coverage Gap
DiseaseThe relationship between disease prevalence and pipeline activity — highlighting potential unmet need
Momentum Score
Therapeutic AreaA therapeutic area or disease-level signal combining recent trial starts, company participation, and overall activity
Freshness
Data Sync Status
Each source has an automated pipeline on the schedule shown. Status reflects how fresh the newest record is relative to that cadence — Current, Delayed, or Stale when a feed has fallen well behind.
| Source | Frequency | Records | Last Data | Status |
|---|---|---|---|---|
| ClinicalTrials.gov | Daily | 235,549 | Jul 9, 2026 | Current · synced today |
| FDA Orange Book | Weekly | 10,007 | Jun 24, 2026 | Failed · synced 4d ago |
| FDA Purple Book | Weekly | 281 | Jun 24, 2026 | Failed · synced 4d ago |
| openFDA | Weekly | 212 | Jul 5, 2026 | Syncing · synced 4d ago |
| PubMed / MEDLINE | Twice weekly | 260 | Jul 7, 2026 | Current · synced 2d ago |
| SEC EDGAR | Daily | 20 | Jun 18, 2026 | Current · synced today |
| Wikidata | Weekly | 138 | Jul 1, 2026 | Current · synced 1d ago |
| Financial Modeling Prep | Weekly | 58 | Jul 1, 2026 | Current · synced 1d ago |