1. What Research Radar tracks
Thousands of rare diseases are recognized in medical classifications, but their research profiles differ markedly. Our initial catalog includes 20 selected rare diseases, chosen for monitoring potential, scientific activity and relevance to patients and research organizations. They were not identified by exhaustively ranking every known rare disease.
The goal is to help readers locate active research, compare its pace, and explore original sources. A high publication count does not establish that a disease is more common, more serious, more successfully treated, or more deserving of attention.
2. Publication volume: available data
For each monitored disease, the local Norqia Digest application sends a disease-specific search to the PubMed database through NCBI E-utilities. A query contains an approved set of disease names and synonyms joined with OR, limited to title and abstract fields. PubMed returns the number of matching records; the dashboard does not infer counts from a capped list of downloaded manuscripts.
Searches use PubMed's publication-date filter (datetype=pdat), with explicit start and end dates. The initial four options are:
| Period | Definition |
|---|---|
| Last 30 days | Thirty inclusive calendar dates ending on the reporting date |
| Last 3 months | Three trailing calendar months |
| Last 6 months | Six trailing calendar months (the default) |
| Last 12 months | Twelve trailing calendar months, not the previous calendar year |
Every published dataset records the exact date windows, original query expressions, query-version identifiers, retrieval timestamps, and snapshot provenance. The reporting calendar uses America/New_York; source timestamps are recorded separately.
What counts include—and exclude
- These are PubMed query matches, not necessarily original research studies. Matching reviews and other indexed publication types may count.
- One paper mentioning two diseases may be counted for both. Totals summed across diseases do not represent distinct articles.
- Search expressions differ by disease. A broad synonym such as “motor neuron disease” may capture related conditions, potentially affecting comparisons.
- Studies without a matching phrase in the title or abstract can be missed. MeSH-only indexing and unrelated full-text terms are not used in this version.
- New or corrected PubMed records can change historical publication-date counts over time. A snapshot therefore represents what PubMed returned at retrieval.
For reproducibility, query revisions are versioned and stored alongside the results rather than silently retroactively modifying historical calculations. Full query definitions are accessible from the individual disease profiles. Additional technical details are available in the NCBI E-utilities documentation.
2A. Patient-population comparisons: under review
Population estimates and publications-per-patient comparisons are not included in the current Research Radar release. Patient-population sources vary in definitions, geography and reference years; reliable denominators must be reviewed before any such metric can be published. For now, Research Radar reports PubMed query matches only.
3. Clinical-trial activity: planned
Research Radar is designed to integrate structured data from ClinicalTrials.gov. Proposed indicators include the number of currently active interventional studies, studies registered over the preceding six months, and studies recruiting participants.
| Trial measure | Proposed share of clinical-trial score |
|---|---|
| Currently active interventional trials | 50% |
| New interventional studies registered in six months | 30% |
| Currently recruiting interventional trials | 20% |
Trials would be deduplicated by NCT identifier. Recruiting studies also appear in the broader active category; that overlap is intentional and disclosed. Observational studies and registries should be visible separately rather than silently pooled with treatment trials. Recruitment and activity statuses are current-state measurements, while new registrations are measured over a dated interval.
4. Public interest and community perspectives: planned
Public information-seeking is a different dimension from scientific publications and clinical development. An initial accessible proxy would be disease-specific readership of English Wikipedia articles through the Wikimedia Analytics API, provided accurate disease-to-article mapping can be validated.
Wikipedia pageviews are not a measure of patient-community needs. They do not identify readers or tell us about lived experiences, insurance barriers, unmet needs, or treatment outcomes. Missing Wikipedia data must not be scored as zero public interest.
A separate Community Perspectives program may eventually analyze permissioned, appropriately de-identified and aggregated discussions and questions. It must respect community authorization, privacy, sampling limitations, and the distinction between reported experiences and scientific evidence.
5. The proposed combined Research Radar Index
Our working model would retain the three source-specific rankings and calculate a fourth composite indicator once all inputs are sufficiently complete and comparable:
| Component | Proposed weight | What it represents |
|---|---|---|
| P — Publications | 40% | Six-month PubMed publication activity |
| T — Clinical trials | 35% | Active, newly registered and recruiting interventional studies |
| C — Public interest | 25% | Validated information-seeking proxy; not a measure of patient needs |
Each component would be transformed onto a comparable 0–100 scale before combination. Logarithmic scaling is proposed for publication counts to limit domination by heavily published conditions. Clinical-trial and public-interest measures require their own validated transformations and reference distributions.
The weights above are editorial starting assumptions, not empirically validated coefficients. We intend to test alternative weightings, normalization methods, data coverage, and sensitivity to rare conditions with smaller research communities. We will not present a combined ranking when a necessary component is unavailable or methodologically unsuitable.
6. Collection, review, and website publishing
- Collect locally. Norqia Digest queries public databases and stores independent, timestamped research snapshots in the local application.
- Validate. Complete datasets must pass consistency checks and preserve disease/query provenance. Failed or partial source refreshes do not replace completed results.
- Review and publish. The editor approves a data-only export containing counts, dates, query expressions and methodology metadata. The public website stores an immutable snapshot and releases an approved version.
- Retain history. Corrections can create a new released revision without deleting earlier snapshots. The last approved public data remains accessible even if the local computer is offline.
Automated descriptive sentences on the public website are calculated deterministically from published numbers. Scientific interpretation, medical implications and qualitative community conclusions require separate editorial review.
Interpret the evidence, not just the chart.
Research Radar aims to make research activity discoverable and transparent. It does not rank the medical importance of rare diseases, assess the quality of individual papers, or provide medical advice.
Return to current Research Radar data →Methodology version: radar-method-v1 · Published Research Radar data retains its original reporting dates, source citations and revision identifiers.