Reclassification Is All You Missed: Surfacing Variant Reinterpretation Need in Online Health Communities at Population Scale
Authors: Kamalakkannan Ravi, Mengshu Nie, Annapurna Poduri, Cong Liu
Organizations: Division of Genetics and Genomics, Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA · Department of Neurology, Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA · Computational Health Informatics Program, Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA
Genetic variants are routinely reclassified in ClinVar after a patient is tested, sometimes from uncertain to pathogenic, yet patient recontact is inconsistent, with no clear duty and responsibility diffused among labs, providers, and patients. Many affected people discuss their results in public online health communities that clinical recontact often does not reach. We built PRISM, an agentic large-language-model pipeline that screens these posts, checks whether the variant was later reclassified, and drafts a disclaimer-bearing outreach message, in a no-contact dry run. Across 3.9 million posts from 32 communities, agent-predicted reinterpretation-relevant discussion is rare (0.288%) but concentrated in specialist genetics communities (6.30% versus 0.10% in general medical-advice communities) and growing fivefold since 2014; PRISM surfaces 123 candidate classification differences, 5 of them from an uncertain submission to pathogenic or likely pathogenic. On a held-out set of 30 real-post drafts, a clinician judged 27 factually grounded and 15 fit to send: factual accuracy is necessary but not sufficient for clinically acceptable outreach. PRISM contributes a population-scale reinterpretation-need map with code and data to be released upon publication.
Figures & tables
Figure 2: Rate and volume dissociate. Each of the 32 communities is positioned by size (posts) against reinterpretation-relevant (B2) rate, both on log scales; point area scales with B2 count, dashed contours mark constant B2 volume, and the dotted line is the corpus-wide rate (0.288%). Specialist genetics communities (purple) sit upper-left (small, high-rate) and general-advice forums (grey) lower-right (large, low-rate). Inset: the B2 rate by year (2026 is a partial year).
From
To
Pool
Dev
Test
VUS
Conflicting
53
8
12
P/LP
Conflicting
40
11
8
B/LB
Conflicting
19
6
4
VUS
P/LP
5
2
3
Conflicting
B/LB
2
1
1
B/LB
VUS
1
0
1
Table 1: The 123-post outreach pool by reclassification type (historical vs current ClinVar classification). VUS = uncertain, P/LP = pathogenic or likely pathogenic, B/LB = benign or likely benign; the VUS-to-P/LP row (potentially highest-consequence) is shaded. Dev and Test are the disjoint prompt-optimization and clinician-review subsets.
Figure 3: Held-out examples: PRISM’s recontact drafts and the clinician’s verdict. Two of the N=30 test-set drafts, one fit, one unfit, paraphrased and excerpted for privacy/Terms-of-Use. Across the held-out set, 27 of 30 stated the reclassification correctly, while 15 of 30 drafts (50%) were rated fit to send under the clinician’s overall call.
Scientific and clinical decision-making depends on evidence from the primary literature, but existing standards for representing that evidence (FHIR Evidence, ECO, SEPIO, and the GA4GH Genomic Knowledge Standards) are oriented toward clinical-trial workflows, evidence codes, or single-variant assertions, and do not capture the fine-grained, domain-specific structure of claims in basic and pre-clinical research. We introduce a semantic model for scientific evidence with three core classes, specialize it for genetics, align it structurally to FHIR Evidence with a SEPIO-anchored credibility decomposition, and attach a compact dimensional vocabulary whose conditional-activation rules are validated by a SHACL schema for the implemented constraints. Using clinical variant interpretation as the driving use case, we evaluate the model through a human-AI annotation pilot over six genetics papers, yielding 28 evidence items and 95 source-anchored assertions, with a workflow that keeps curator-authored reference annotations distinct from AI-drafted annotations. Treating the pilot as a feasibility study rather than a benchmark, we argue that the model is a useful increment toward trustworthy, AI-ready infrastructure for variant interpretation: a reference data model and validation schema for representing genetic evidence.
Michael Bouzinier, Dmitry Etin
Harvard University, Cambridge, MA, USA · IDEXX Laboratories, Westbrook, ME, USA · Forome Association, Newton, MA, USA +1
People share mental health diagnoses on social media, yet how such language becomes visible around their self-disclosure, and whether community engagement tracks it, remain unexamined across conditions. We analyze 89,605 Reddit posts from 739 users across eight conditions, removing each user's diagnosis disclosure and aligning their surrounding posts to that anchor. Within the pre-disclosure year, language-visible burden was highest in the month before disclosure for six conditions, earlier for post-traumatic stress disorder and furthest from it for borderline personality disorder, and remained visible afterward rather than resolving. The theme Seeking Clinical Explanations showed the largest early-to-late difference before disclosure in five conditions, yet engagement rarely tracked what users wrote: only 9 of 360 language--engagement correlations survived correction. Disclosure is therefore a waypoint in an unevenly visible process, and we offer implications for community practice and platform design where engagement metrics do not reflect clinical need.
Renkai Ma, Lingyao Li, Shanting Chen +3
College of Information Science University of Arizona Tucson, Arizona, USA · Department of Computer and Information Science and Engineering University of Florida Gainesville, Florida, USA
Disease severity classification for genetic conditions is subjective and labor-intensive, creating bottlenecks in genomic screening, where commercial panels vary widely in size and overlap. We developed an autonomous AI agent integrating Reasoning and Acting (ReAct) with Retrieval-Augmented Generation (RAG) to classify 10,211 Human Phenotype Ontology terms. It uses American College of Medical Genetics (ACMG)-endorsed severity guidelines and American College of Obstetricians and Gynecologists (ACOG) quality-of-life criteria to retrieve PubMed literature, generate interpretable reasoning chains, and independently verify claims. At the phenotype level, using expert-curated cohorts, the agent achieved 93.55% accuracy (MCC 0.9237) with 82.6% to 91.4% of claims supported by direct evidence or valid inferences. Gene-level severity was aggregated across 8,738 pairs, identifying 3,283 autosomal recessive pairs with severe or profound presentations. External validation showed 95.2% concordance with Mackenzie's Mission gene list. This system enables standardized panel design by providing reliable, automated classification supported by direct evidence.
Tohid Ghasemnejad, Ahmadreza Argha, Mark Grosser +7
UNSW BioMedical Machine Learning Lab (BML), School of Biomedical Engineering, UNSW Sydney, Sydney, NSW 2052, Australia · School of Biomedical Engineering, UNSW Sydney, NSW 2052, Australia · 23Strands, Pyrmont, Australia +8