The undiagnosed diseases program: Approach to diagnosis

Identifying and Extracting Rare Diseases and Their Phenotypes with Large Language Models

Purpose

Phenotyping is critical for informing rare disease diagnosis and treatment, but disease phenotypes are often embedded in unstructured text. While natural language processing (NLP) can automate extraction, a major bottleneck is developing annotated corpora. Recently, prompt learning with large language models (LLMs) has been shown to lead to generalizable results without any (zero-shot) or few annotated samples (few-shot), but none have explored this for rare diseases. Our work is the first to study prompt learning for identifying and extracting rare disease phenotypes in the zero- and few-shot settings.

Methods

We compared the performance of prompt learning with ChatGPT and fine-tuning with BioClinicalBERT. We engineered novel prompts for ChatGPT to identify and extract rare diseases and their phenotypes (e.g., diseases, symptoms, and signs), established a benchmark for evaluating its performance, and conducted an in-depth error analysis.

Results

Overall, fine-tuning BioClinicalBERT resulted in higher performance (F1 of 0.689) than ChatGPT (F1 of 0.472 and 0.610 in the zero- and few-shot settings, respectively). However, ChatGPT achieved higher accuracy for rare diseases and signs in the one-shot setting (F1 of 0.778 and 0.725). Conversational, sentence-based prompts generally achieved higher accuracy than structured lists.

Conclusion

Prompt learning using ChatGPT has the potential to match or outperform fine-tuning BioClinicalBERT at extracting rare diseases and signs with just one annotated sample. Given its accessibility, ChatGPT could be leveraged to extract these entities without relying on a large, annotated corpus. While LLMs can support rare disease phenotyping, researchers should critically evaluate model outputs to ensure phenotyping accuracy.

International Undiagnosed Diseases Programs (UDPs): components and outcomes

Over the last 15 years, Undiagnosed Diseases Programs have emerged to address the significant number of individuals with suspected but undiagnosed rare genetic diseases, integrating research and clinical care to optimize diagnostic outcomes. This narrative review summarizes the published literature surrounding Undiagnosed Diseases Programs worldwide, including thirteen studies that evaluate outcomes and two commentary papers. Commonalities in the diagnostic and research process of Undiagnosed Diseases Programs are explored through an appraisal of available literature. This exploration allowed for an assessment of the strengths and limitations of each of the six common steps, namely enrollment, comprehensive clinical phenotyping, research diagnostics, data sharing and matchmaking, results, and follow-up. Current literature highlights the potential utility of Undiagnosed Diseases Programs in research diagnostics. Since participants have often had extensive previous genetic studies, research pipelines allow for diagnostic approaches beyond exome or whole genome sequencing, through reanalysis using research-grade bioinformatics tools and multi-omics technologies. The overall diagnostic yield is presented by study, since different selection criteria at enrollment and reporting processes make comparisons challenging and not particularly informative. Nonetheless, diagnostic yield in an undiagnosed cohort reflects the potential of an Undiagnosed Diseases Program. Further comparisons and exploration of the outcomes of Undiagnosed Diseases Programs worldwide will allow for the development and improvement of the diagnostic and research process and in turn improve the value and utility of an Undiagnosed Diseases Program.

Diagnosis and discovery: Insights from the NIH Undiagnosed Diseases Program

Living with an undiagnosed medical condition places a tremendous burden on patients, their families, and their healthcare providers. The Undiagnosed Diseases Program (UDP) was established at the National Institutes of Health (NIH) in 2008 with the primary goals of providing a diagnosis for patients with mysterious conditions and advancing medical knowledge about rare and common diseases. The program reviews applications from referring clinicians for cases that are considered undiagnosed despite a thorough evaluation. Those that are accepted receive clinical evaluations involving deep phenotyping and genetic testing that includes exome and genomic sequencing. Selected candidate gene variants are evaluated by collaborators using functional assays. Since its inception, the UDP has received more than 4500 applications and has completed evaluations on nearly 1300 individuals. Here we present six cases that exemplify the discovery of novel disease mechanisms, the importance of deep phenotyping for rare diseases, and how genetic diagnoses have led to appropriate treatment. The creation of the Undiagnosed Diseases Network (UDN) in 2014 has substantially increased the number of patients evaluated and allowed for greater opportunities for data sharing. Expansion to the Undiagnosed Diseases Network International (UDNI) has the possibility to extend this reach even farther. Together, networks of undiagnosed diseases programs are powerful tools to advance our knowledge of pathophysiology, accelerate accurate diagnoses, and improve patient care for patients with rare conditions.

Adults with lysosomal storage diseases in the undiagnosed diseases network

OBJECTIVES

To review the referral and clinical characteristics of adult patients diagnosed with lysosomal storage diseases (LSD) through the Undiagnosed Diseases Network (UDN).

METHODS

Retrospective review of both application and evaluation records for adults admitted to the UDN with a final diagnosis of a lysosomal storage disease.

RESULTS

Ten patients were identified. Final diagnoses included late onset Tay Sachs, attenuated MPS I, MPS IIIA, MPS IIIB, and MPS IIIC. Most patients presented with neurocognitive changes. Prior to referral, all patients had been evaluated by neurology, four patients underwent phenotype specific panel testing that did not include the causative gene, and four patients had non-diagnostic clinical exome sequencing.

CONCLUSIONS

LSDs figure highly in the differential diagnosis of neurometabolic disorders in pediatric onset progressive diseases. In adults, their subtle initial presentations overlap with symptoms of more common disorders and less practitioner awareness may lead to prolonged diagnostic challenges.

Seven years since the launch of the Matchmaker Exchange: The evolution of genomic matchmaking

The Matchmaker Exchange (MME) was launched in 2015 to provide a robust mechanism to discover novel disease-gene relationships. It operates as a federated network connecting databases holding relevant data using a common application programming interface, where two or more users are looking for a match for the same gene (two-sided matchmaking). Seven years from its launch, it is clear that the MME is making outstanding contributions to understanding the morbid anatomy of the genome. The number of unique genes present across the MME has steadily increased over time; there are currently >13,520 unique genes (~68% of all protein-coding genes) connected across the MME's eight genomic matchmaking nodes, GeneMatcher, DECIPHER, PhenomeCentral, MyGene2, seqr, Initiative on Rare and Undiagnosed Disease, PatientMatcher, and the RD-Connect Genome-Phenome Analysis Platform. The collective data set accessible across the MME currently includes more than 120,000 cases from over 12,000 contributors in 98 countries. The discovery of potential new disease-gene relationships is happening daily and international collaborative teams are moving these advances forward to publication, now numbering well over 500. Expansion of data sharing into routine clinical practice by clinicians, genetic counselors, and clinical laboratories has ensured access to discovery for even more individuals with undiagnosed rare genetic diseases. Tens of thousands of patients and their family members have been directly or indirectly impacted by the discoveries facilitated by two-sided genomic matchmaking. MME supports further connections to the literature (PubCaseFinder) and to human and model organism resources (Monarch Initiative) and scientists (ModelMatcher). Efforts are now underway to explore additional approaches to matchmaking at the gene or variant level where there is only one querier (one-sided matchmaking). Genomic matchmaking has proven its utility over the past 7 years and will continue to facilitate discoveries in the years to come.

Diagnostic journeys: characterization of patients and diagnostic outcomes from an academic second opinion clinic

OBJECTIVES

Diagnostic programs and second opinion clinics have grown and evolved in the recent years to help patients with rare, puzzling, and complex conditions who often suffer prolonged diagnostic journeys, but there is a paucity of literature on the clinical characteristics of these patients and the efficacy of these diagnostic programs. This study aims to characterize the diagnostic journey, case features, and diagnostic outcomes of patients referred to a team-based second opinion clinic at Stanford.

METHODS

Retrospective chart review was performed for 237 patients evaluated for diagnostic second opinion in the Stanford Consultative Medicine Clinic over a 5 year period. Descriptive case features and diagnostic outcomes were assessed, and correlation between the two was analyzed.

RESULTS

Sixty-three percent of our patients were women. 49% of patients had a potential precipitating event within about a month prior to the start of their illness, such as medication change, infection, or medical procedure. A single clear diagnosis was determined in 33% of cases, whereas the remaining cases were assessed to have multifactorial contributors/diagnoses (20%) or remained unclear despite extensive evaluation (47%). Shorter duration of illness, fewer prior specialties seen, and single chief symptom were associated with higher likelihood of achieving a single clear diagnosis.

CONCLUSIONS

A single-site academic consultative service can offer additional diagnostic insights for about half of all patients evaluated for puzzling conditions. Better understanding of the clinical patterns and patient experiences gained from this study helps inform strategies to shorten their diagnostic odysseys.

Advances in the development of PubCaseFinder, including the new application programming interface and matching algorithm

Over 10,000 rare genetic diseases have been identified, and millions of newborns are affected by severe rare genetic diseases each year. A variety of Human Phenotype Ontology (HPO)-based clinical decision support systems (CDSS) and patient repositories have been developed to support clinicians in diagnosing patients with suspected rare genetic diseases. In September 2017, we released PubCaseFinder (https://pubcasefinder.dbcls.jp), a web-based CDSS that provides ranked lists of genetic and rare diseases using HPO-based phenotypic similarities, where top-listed diseases represent the most likely differential diagnosis. We also developed a Matchmaker Exchange (MME) application programming interface (API) to query PubCaseFinder, which has been adopted by several patient repositories. In this paper, we describe notable updates regarding PubCaseFinder, the GeneYenta matching algorithm implemented in PubCaseFinder, and the PubCaseFinder API. The updated GeneYenta matching algorithm improves the performance of the CDSS automated differential diagnosis function. Moreover, the updated PubCaseFinder and new API empower patient repositories participating in MME and medical professionals to actively use HPO-based resources. This article is protected by copyright. All rights reserved.

Strategies to Uplift Novel Mendelian Gene Discovery for Improved Clinical Outcomes

Rare genetic disorders, while individually rare, are collectively common. They represent some of the most severe disorders affecting patients worldwide with significant morbidity and mortality. Over the last decade, advances in genomic methods have significantly uplifted diagnostic rates for patients and facilitated novel and targeted therapies. However, many patients with rare genetic disorders still remain undiagnosed as the genetic etiology of only a proportion of Mendelian conditions has been discovered to date. This article explores existing strategies to identify novel Mendelian genes and how these discoveries impact clinical care and therapeutics. We discuss the importance of data sharing, phenotype-driven approaches, patient-led approaches, utilization of large-scale genomic sequencing projects, constraint-based methods, integration of multi-omics data, and gene-to-patient methods. We further consider the health economic advantages of novel gene discovery and speculate on potential future methods for improved clinical outcomes.

Commonalities across computational workflows for uncovering explanatory variants in undiagnosed cases

PURPOSE

Genomic sequencing has become an increasingly powerful and relevant tool to be leveraged for the discovery of genetic aberrations underlying rare, Mendelian conditions. Although the computational tools incorporated into diagnostic workflows for this task are continually evolving and improving, we nevertheless sought to investigate commonalities across sequencing processing workflows to reveal consensus and standard practice tools and highlight exploratory analyses where technical and theoretical method improvements would be most impactful.

METHODS

We collected details regarding the computational approaches used by a genetic testing laboratory and 11 clinical research sites in the United States participating in the Undiagnosed Diseases Network via meetings with bioinformaticians, online survey forms, and analyses of internal protocols.

RESULTS

We found that tools for processing genomic sequencing data can be grouped into four distinct categories. Whereas well-established practices exist for initial variant calling and quality control steps, there is substantial divergence across sites in later stages for variant prioritization and multimodal data integration, demonstrating a diversity of approaches for solving the most mysterious undiagnosed cases.

CONCLUSION

The largest differences across diagnostic workflows suggest that advances in structural variant detection, noncoding variant interpretation, and integration of additional biomedical data may be especially promising for solving chronically undiagnosed cases.