TrialNavigator

I began this project the same way I start most of them: with a literature search to understand what already existed. The question was simple enough: is there an automated trial matching tool for CACNA1A patients?

Nope. There are OpenFDA API wrappers, commercial pharmacovigilance platforms, and general-purpose clinical trial aggregators aimed at researchers. The CACNA1A Foundation has a trials page, but it's manually maintained. There is no tool that continuously monitors ClinicalTrials.gov for CACNA1A-related studies, translates criteria, maps sites, and delivers updates to families via email.

The ClinicalTrials.gov API (REST v2) is free, requires no authentication, and returns well-structured JSON for up to 100 results per query. Running four queries (across CACNA1A, familial hemiplegic migraine, episodic ataxia type 2, and spinocerebellar ataxia type 6) and deduplicating on NCT ID gives a comprehensive view of the current trial landscape. As of April 2026, this covers several dozen trials across phases and statuses.

Understanding this landscape took a bit of time. Phase I trials test safety in small populations. Phase II adds preliminary efficacy data. Phase III trials are the large, randomized, controlled studies that regulatory agencies use to make approval decisions. NCT07221292 is Phase III, which means it has already cleared the safety and preliminary efficacy bar established by IntraBio's earlier work in Niemann-Pick type C and Ataxia-Telangiectasia. For a patient family, that matters considerably: you're not enrolling in something with an unknown safety profile.

The other thing worth understanding about ClinicalTrials.gov data is what it doesn't tell you. Trial records are submitted and updated by sponsors, not by the registry itself. A trial listed as "Recruiting" may be actively enrolling or may have a frozen enrollment that hasn't been updated in the registry. Contact information for sites is often incomplete. The start date listed may be an estimate. None of this makes the data unusable, but it means the right call to action is always "contact the trial coordinator directly" rather than "you are eligible, proceed." TrialNavigator is built around this principle: it lowers the barrier to finding the right conversation, not to replacing it.

The frontend is React, deployed on Vercel. The backend is Python with SQLite, running on a GitHub Actions weekly cron. No paid infrastructure is required; the total cost at the scale TrialNavigator will operate is zero dollars per month.

The ClinicalTrials.gov REST API v2 is documented at clinicaltrials.gov/data-api/api. The base query endpoint is /api/v2/studies with query parameters for condition and keyword. Each response contains the trial's identification module (NCT ID, title), status module (overall status, dates), eligibility module (free-text criteria, age range, sex), contacts and locations module (site names, cities, contact emails), and description module (brief summary, detailed description). The updater.py script runs all four queries, deduplicates on NCT ID, and stores normalized records in SQLite. It also diffs against the previous week's records to generate a change digest: new trials, status changes, and new site additions. This digest is written to public/update_digest.json and committed back to the GitHub repo via the Actions workflow, where the frontend can serve it to subscribed users.

Geographic distance is a practically significant feature in rare disease trial matching. For conditions affecting tens of thousands of patients nationally, the nearest enrolling site might be five hundred miles away. TrialNavigator computes site distances using the Haversine formula, which gives great-circle distance between two latitude/longitude points on a sphere:

a = sin²(Δlat/2) + cos(lat₁)·cos(lat₂)·sin²(Δlon/2) d = 2R·arcsin(√a)

where R = 3958.8 miles (Earth's mean radius). This is implemented from scratch in src/utils/geo.js. User zip codes are geocoded to approximate coordinates using the US Census Geocoder, which is free and covers all US zip codes. The driving time estimate is a rough heuristic: approximately 60 mph for distances over 50 miles. It is labeled as an estimate in the UI. Sites are sorted by this distance. For rare disease trials, travel support is often available and worth asking about; the site detail view flags this where disclosed in the trial record.

The plain-language translation of eligibility criteria is the other technically interesting backend component. Raw ClinicalTrials.gov eligibility text is written in regulatory language: "Participants who have received prior treatment with an investigational CACNA1A-targeted therapy within 90 days of screening are excluded." The system prompt for translation is: "Translate the following clinical trial eligibility criterion into plain language that a patient or family can understand without medical training. Be specific and concrete. Maximum 2 sentences." This is called per-criterion via the Claude API (Haiku model, for cost efficiency at scale), and both the original and translated versions are stored. If no API key is configured, the original criterion text is displayed as a fallback.

The eligibility checker is a five-step wizard: age, confirmed diagnosis, CACNA1A variant, current medications, and results. The assessment logic runs in src/utils/eligibility.js.

For each active recruiting trial, three criteria are checked sequentially. Age is assessed against the trial's minimum and maximum age fields extracted from the eligibility module. Diagnosis is a user self-report; if the answer is No, the trial receives a RED indicator and the user is directed to pursue genetic confirmation as a prerequisite. Variant matching uses a lookup table derived from published CACNA1A variant classifications: R192Q, S218L, T666M, and others map to FHM1 (gain-of-function); G293R, F1490K and others map to EA2 (loss-of-function). If the user's variant maps to a phenotype explicitly listed in the trial's conditions, the indicator is GREEN. If the variant is unknown, the indicator is YELLOW with a note to contact the coordinator. If there is a mismatch, it is RED.

The traffic-light framework is deliberate. Clinical trial eligibility assessment requires a physician's judgment and a full review of the trial protocol; TrialNavigator cannot and should not replicate that. What it can do is give a patient a starting point: "Based on your age and R192Q variant, this trial looks like it might be worth a call." A RED result does not mean you are ineligible; it means one criterion on its face looks like a mismatch. The system explicitly avoids framing results as screening-out decisions, and the UI instruction for RED and YELLOW results is consistently "contact the trial coordinator to discuss eligibility."

The variant lookup table is a simplified version of what a full TraitMap integration would provide. Once TraitMap is complete within the MiSOF framework, TrialNavigator can be updated to pull phenotype classification dynamically rather than from a hard-coded lookup. The current table covers the most clinically significant FHM1 GoF variants and the most common EA2 LoF variants; it is documented in src/utils/eligibility.js and straightforward to extend.

N-Acetyl-L-Leucine (levacetylleucine, IB1001) is a modified amino acid developed by IntraBio Inc. It has completed pivotal Phase III trials for Niemann-Pick disease type C and Ataxia-Telangiectasia, both with positive results and established safety profiles. In February 2026, the EMA issued a positive opinion recommending orphan medicinal product designation for levacetylleucine for CACNA1A disorders. The Phase III CACNA1A trial (NCT07221292) is a multinational, randomized, double-blind, placebo-controlled crossover study enrolling patients age 4 and older.

The mechanism is investigational but biologically plausible for FHM1. GoF CACNA1A mutations (R192Q being the most studied) increase calcium influx through Cav2.1 channels, contributing to neuronal hyperexcitability and downstream cortical spreading depression. Levacetylleucine is proposed to improve lysosomal calcium sequestration, reducing intracellular calcium dysregulation. Whether this translates to clinical benefit in FHM1 is what the trial is designed to determine; the prior positive results in two other neurological excitability disorders justify the hypothesis.

The crossover design is clinically and practically important for patient families. In a traditional randomized controlled trial, half the participants receive placebo for the entire study duration. In a crossover design, each participant receives both active drug and placebo in sequence, separated by a washout period. For NCT07221292, this means two 12-week periods: everyone receives levacetylleucine at some point. The trial is also compatible with current medications, meaning patients who are already on topiramate, amitriptyline, or other preventives can enroll without discontinuing their regimen. The route of administration is oral: a granule sachet dissolved in water or juice. These features substantially lower the practical barrier to participation.

TrialNavigator's dedicated page for NCT07221292 includes plain-language answers to the questions families actually ask: can I stay on my medications, what if I get placebo, is this a cure, how do I express interest before enrollment opens. The contact for IntraBio is ccfields@intrabio.com. Once TrialNavigator is deployed, IntraBio will receive a notification that an independent patient-facing tool is directing CACNA1A families to their trial. This costs nothing and puts the project on their radar for future collaboration.

Building the tool is the easier half. Getting it in front of the families who need it is the other half, and the CACNA1A Foundation is the only realistic path to that at scale.

The Foundation's patient community is the largest organized group of CACNA1A families that exists. They have conference attendees, natural history study enrollees, an email list, and researcher relationships. Their current trial listing is a manually maintained webpage. TrialNavigator offers them something better: an automated, continuously updated, patient-readable tool that replaces their manual listing and improves on it substantially. 

The ask is specific: embed TrialNavigator on cacna1a.org, or link to it prominently from the trials section. The tool is open-source on GitHub (axshoe/trialnavigator), free to use, and requires no maintenance from the Foundation. 

The broader observation about what this kind of tool represents is worth stating plainly. Every methodological piece of TrialNavigator was available before I built it. The ClinicalTrials.gov API has existed since 2012. Haversine distance calculation has a closed-form solution. Plain-language translation is a standard API call. The variant-to-phenotype lookup is a table that anyone with access to the FHM1 literature could construct in an afternoon. The contribution is not technical novelty. It is that someone connected these pieces together, pointed them at a specific patient population, and made the result publicly accessible and automatically maintained. In rare disease research, this kind of infrastructure work matters in proportion not to its computational sophistication but to how many people it reaches.

The original TrialNavigator was built for one purpose: to make sure CACNA1A families knew NCT07221292 existed before enrollment opened in September 2026. It was a disease-specific tool, and a deliberately narrow one. The architecture happened to be disease-agnostic from the start (the ClinicalTrials.gov API takes any condition query), but the disease registry, the variant matching module, and the featured trial logic were all CACNA1A-specific.

My conversation with Sunitha Malepati at the CACNA1A Foundation and the Buffalo Initiative changed the scope of my thinking. The Buffalo Initiative pipeline tracker covers 62 patient-led drug development programs across rare neurological, metabolic, and neuromuscular conditions. Every one of those programs will eventually need enrollment infrastructure. Most of those patient organizations currently maintain manually-updated trial listings on their websites, a process that requires someone to remember to check ClinicalTrials.gov, understand what they're looking at, and update a page: none of which is guaranteed to happen on any particular schedule. The gap is structural, not individual.

Phase 6 was the process of turning a CACNA1A tool into infrastructure that any patient organization in the Buffalo network can embed with one line of HTML.

What changed technically. The disease registry in v2 covers 26 conditions drawn from the Buffalo Initiative pipeline, each with curated ClinicalTrials.gov search queries that go substantially beyond single gene-name searches. Many trials for rare genetic conditions are registered under umbrella terms: "developmental epileptic encephalopathy" instead of STXBP1, "bexicaserin" or "LP352" as the intervention name rather than the condition name, NCT IDs included directly as search terms for known trials that might not surface otherwise. Casting multiple overlapping queries per disease and deduplicating on NCT ID catches what single-term searches miss.

The sorting logic was overhauled. Opening-soon and actively recruiting trials appear at the top. Completed, terminated, and withdrawn trials appear at the bottom, greyed out to 55% opacity. This matters more than it sounds: a family in crisis should not have to scroll past a completed 2019 study to find out whether anything is currently enrolling. A status legend explains each category in plain terms, i.e. "Enrolling now: accepting new participants" rather than the API's "RECRUITING."

The embed system is the substantive new capability. Any patient organization can add a filtered view of TrialNavigator to their own website by adding a single iframe tag pointing to the deployed URL with ?embed=1&disease=their_condition. The navigation, footer, and disease browser disappear in embed mode. They get a branded, automatically-updating trial listing that requires zero maintenance on their end. The CACNA1A Foundation version and the Buffalo Initiative version are the same deployed application, pointed at different URL parameters.

Working through the full disease list was frustration-inducing experience. Here are a few examples: KIF1A-associated neurological disorder has no interventional treatment trials exist as of 2025. KCNH1-related disorders have no trials of any kind. FRRS1L disorder has fewer than fifty reported patients worldwide and no registry. For these conditions, TrialNavigator can only return observational and natural history studies alongside a plain-language note explaining why interventional results are absent. This is, by all unfortunate means, the real state of a field where 95% of rare diseases have no FDA-approved treatment and most are years from their first clinical trial. 

That context is the reason tools like TrialNavigator exist. The gap between a trial opening and a patient finding it is not the only relevant gap. The larger gap is between conditions that have drug development programs at all and the thousands that do not. TrialNavigator is infrastructure for the ones that do. For the ones that don't, the most useful thing it can do is say so clearly, and point families toward natural history studies that are building the foundation future trials will stand on.

The v3 direction is as follows: the current v2 TrialNavigator disease registry is hand-curated from the Buffalo Initiative pipeline. NORD maintains a searchable database covering 1,200 rare diseases, and has over 340 patient advocacy organizations as members. Roughly 10,000 rare diseases are estimated to exist, with only about 5% having any FDA-approved treatment. A fully disease-agnostic version of TrialNavigator would pull disease definitions and associated search terms from Orphanet's classification API or NORD's database, auto-generate ClinicalTrials.gov queries, and make every condition searchable without manual curation. The embed system would remain, and any organization could point the iframe at their condition by name or OMIM code. This is a different scale of engineering problem, but not really a different architecture.

The original TrialNavigator question was a bit narrow-minded: how do CACNA1A families find out that NCT07221292 exists before enrollment opens in September 2026? Phase 6 expanded that question to the Buffalo Initiative's 26 conditions. Phase 7 asks whether this same infrastructure can be extrapolated to the entire rare disease space.

The answer is yes, absolutely, and with important qualifications. There are approximately 10,000 rare diseases, of which roughly 7,000 have been formally characterized and catalogued in databases like Orphanet. Only about 5% have FDA-approved treatments. The universe of patient advocacy organizations with active presences is somewhere between 600 and 1,500 in the US alone, with Global Genes' Global Advocacy Alliance representing over 820 member organizations globally. NORD has over 340 member organizations.

Every one of these organizations faces the same infrastructure problem, wherein a large chunk of them still maintain manually-updated trial listings on their websites, which require a staff member to periodically check ClinicalTrials.gov, parse what they find, and update a page. This breaks when staff changes, when a trial status changes mid-cycle, or when a new trial opens in the two weeks between updates. For a disease where a Phase III trial is the most important thing happening in the community, a 6/7-week lag in awareness can meaningfully affect enrollment.

RareTrials is the response to that problem, but scaled. The architecture is pretty much the same as TrialNavigator: ClinicalTrials.gov API v2, disease-specific query sets, weekly GitHub Actions update, Vercel deployment, and an embed system that lets any organization add a filtered view to their website with one HTML line. The differences are in scope and design language.

The disease registry in RareTrials v1 covers 4,128 conditions across 18 therapeutic categories, drawn from Orphanet classification, NORD's database, and the Global Genes RARE list. These are organized by category: neurological, metabolic, lysosomal storage, mitochondrial, neuromuscular, hematologic, immunologic, cardiac, connective tissue, pulmonary, renal, dermatologic, ophthalmic, oncologic, endocrine, hepatic, and undiagnosed. Each disease has a set of search queries designed to surface trials registered under gene names, drug names, umbrella terms, and specific NCT IDs for known ongoing studies.

The design language was deliberately much different from TrialNavigator. RareTrials uses Libre Baskerville serif for headings, the same typeface family NIH uses for formal publications, paired with DM Sans for interface elements. The color palette is deep navy, warm cream, and amber, which I chose for its authoritative medical feel rather than tech product. I also decided to integrate 3 stats in the hero section before the search box: 4,128 diseases covered, 10,000+ rare diseases exist, 95% have no approved treatment. Shocking statistics, but its scale is important for families using the tool to know. 

The reason I built RareTrials (technically TrialNavigator v3) separately from TrialNavigator is because of the audience. TrialNavigator is infrastructure for specific partner organizations and is something VP Sunitha Malepati can give to her network, that the CACNA1A Foundation can embed on cacna1a.org, and that each Buffalo Initiative organization gets as a white-labeled view. RareTrials is something any family can find by searching, any organization can embed without a prior relationship, and any researcher can point a patient to when a trial opens and not specific to just the rare diseases associated with the Buffalo Initiative or CACNA1A Foundation. 

The deployment and outreach strategy for RareTrials will also be different from TrialNavigator. I would much rather, that is, not manually contact 500 organizations individually. My circadian rhythm has been utterly demolished this past year (I am typing this at 2:59 AM on a Saturday night). Thus, the right move will be 2 umbrella relationships: registering with NORD as an official patient resource, and contacting Global Genes' Global Advocacy Alliance team, whose 820 member organizations collectively represent most of the organizations worth reaching.