Shortcut learning in medical AI hinders generalization: method for estimating AI model generalization without external data.

Healthcare datasets are becoming larger and more complex, necessitating the development of accurate and generalizable AI models for medical applications. Unstructured datasets, including medical imaging, electrocardiograms, and natural language data, are gaining attention with advancements in deep convolutional neural networks and large language models. However, estimating the generalizability of these models to new healthcare settings without extensive validation on external data remains challenging. In experiments across 13 datasets including X-rays, CTs, ECGs, clinical discharge summaries, and lung auscultation data, our results demonstrate that model performance is frequently overestimated by up to 20% on average due to shortcut learning of hidden data acquisition biases (DAB). Shortcut learning refers to a phenomenon in which an AI model learns to solve a task based on spurious correlations present in the data as opposed to features directly related to the task itself. We propose an open source, bias-corrected external accuracy estimate, PEst, that better estimates external accuracy to within 4% on average by measuring and calibrating for DAB-induced shortcut learning.

Clinical Deployment of Machine Learning Tools in Transplant Medicine: What Does the Future Hold?

Medical applications of machine learning (ML) have shown promise in analyzing patient data to support clinical decision-making and provide patient-specific outcomes. In transplantation, several applications of ML exist which include pretransplant: patient prioritization, donor-recipient matching, organ allocation, and posttransplant outcomes. Numerous studies have shown the development and utility of ML models, which have the potential to augment transplant medicine. Despite increasing efforts to develop robust ML models for clinical use, very few of these tools are deployed in the healthcare setting. Here, we summarize the current applications of ML in transplant and discuss a potential clinical deployment framework using examples in organ transplantation. We identified that creating an interdisciplinary team, curating a reliable dataset, addressing the barriers to implementation, and understanding current clinical evaluation models could help in deploying ML models into the transplant clinic setting.

Use of artificial intelligence for decision-support to avoid high-risk behaviors during laparoscopic cholecystectomy.

INTRODUCTION: Bile duct injuries (BDIs) are a significant source of morbidity among patients undergoing laparoscopic cholecystectomy (LC). GoNoGoNet is an artificial intelligence (AI) algorithm that has been developed and validated to identify safe ("Go") and dangerous ("No-Go") zones of dissection during LC, with the potential to prevent BDIs through real-time intraoperative decision-support. This study evaluates GoNoGoNet's ability to predict Go/No-Go zones during LCs with BDIs. METHODS AND PROCEDURES: Eleven LC videos with BDI (BDI group) were annotated by GoNoGoNet. All tool-tissue interactions, including the one that caused the BDI, were characterized in relation to the algorithm's predicted location of Go/No-Go zones. These were compared to another 11 LC videos with cholecystitis (control group) deemed to represent "safe cholecystectomy" by experts. The probability threshold of GoNoGoNet annotations were then modulated to determine its relationship to Go/No-Go predictions. Data is shown as % difference [99% confidence interval]. RESULTS: Compared to control, the BDI group showed significantly greater proportion of sharp dissection (+ 23.5% [20.0-27.0]), blunt dissection (+ 32.1% [27.2-37.0]), and total interactions (+ 33.6% [31.0-36.2]) outside of the Go zone. Among injury-causing interactions, 4 (36%) were in the No-Go zone, 2 (18%) were in the Go zone, and 5 (45%) were outside both zones, after maximizing the probability threshold of the Go algorithm. CONCLUSION: AI has potential to detect unsafe dissection and prevent BDIs through real-time intraoperative decision-support. More work is needed to determine how to optimize integration of this technology into the operating room workflow and adoption by end-users.

What is the educational value and clinical utility of artificial intelligence for intraoperative and postoperative video analysis? A survey of surgeons and trainees.

INTRODUCTION: Surgical complications often occur due to lapses in judgment and decision-making. Advances in artificial intelligence (AI) have made it possible to train algorithms that identify anatomy and interpret the surgical field. These algorithms can potentially be used for intraoperative decision-support and postoperative video analysis and feedback. Despite the very early success of proof-of-concept algorithms, it remains unknown whether this innovation meets the needs of end-users or how best to deploy it. This study explores users' opinion on the value, usability and design for adapting AI in operating rooms. METHODS: A device-agnostic web-accessible software was developed to provide AI inference either (1) intraoperatively on a live video stream (synchronous mode), or (2) on an uploaded video or image file (asynchronous mode) postoperatively for feedback. A validated AI model (GoNoGoNet), which identifies safe and dangerous zones of dissection during laparoscopic cholecystectomy, was used as the use case. Surgeons and trainees performing laparoscopic cholecystectomy interacted with the AI platform and completed a 5-point Likert scale survey to evaluate the educational value, usability and design of the platform. RESULTS: Twenty participants (11 surgeons and 9 trainees) evaluated the platform intraoperatively (n = 10) and postoperatively (n = 11). The majority agreed or strongly agreed that AI is an effective adjunct to surgical training (81%; neutral = 10%), effective for providing real-time feedback (70%; neutral = 20%), postoperative feedback (73%; neutral = 27%), and capable of improving surgeon confidence (67%; neutral = 29%). Only 40% (neutral = 50%) and 57% (neutral = 43%) believe that the tool is effective in improving intraoperative decisions and performance, or beneficial for patient care, respectively. Overall, 38% (neutral = 43%) reported they would use this platform consistently if available. The majority agreed or strongly agreed that the platform was easy to use (81%; neutral = 14%) and has acceptable resolution (62%; neutral = 24%), while 30% (neutral = 20%) reported that it disrupted the OR workflow, and 20% (neutral = 0%) reported significant time lag. All respondents reported that such a system should be available "on-demand" to turn on/off at their discretion. CONCLUSIONS: Most found AI to be a useful tool for providing support and feedback to surgeons, despite several implementation obstacles. The study findings will inform the future design and usability of this technology in order to optimize its clinical impact and adoption by end-users.

Clinical Implementation of MetaFusion for Accurate Cancer-Driving Fusion Detection from RNA Sequencing.

Oncogenic fusion genes may be identified from next-generation sequencing data, typically RNA-sequencing. However, in a clinical setting, identifying these alterations is challenging against a background of nonrelevant fusion calls that reduce workflow precision and specificity. Furthermore, although numerous algorithms have been developed to detect fusions in RNA-sequencing, there are variations in their individual sensitivities. Here this problem was addressed by introducing MetaFusion into clinical use. Its utility was illustrated when applied to both whole-transcriptome and targeted sequencing data sets. MetaFusion combines ensemble fusion calls from eight individual fusion-calling algorithms with practice-informed identification of gene fusions that are known to be clinically relevant. In doing so, it allows oncogenic fusions to be identified with near-perfect sensitivity and high precision and specificity, significantly outperforming the individual fusion callers it uses as well as existing clinical-grade software. MetaFusion enhances clinical yield over existing methods and is able to identify fusions that have patient relevance for the purposes of diagnosis, prognosis, and treatment.

Conserved transcriptional programming across sex and species after peripheral nerve injury predicts treatments for neuropathic pain.

BACKGROUND AND PURPOSE: Chronic pain is a devastating problem affecting one in five individuals around the globe, with neuropathic pain the most debilitating and poorly treated type of chronic pain. Advances in transcriptomics have contributed to cataloguing diverse cellular pathways and transcriptomic alterations in response to peripheral nerve injury but have focused on phenomenology and classifying transcriptomic responses. EXPERIMENTAL APPROACH: To identifying new types of pain-relieving agents, we compared transcriptional reprogramming changes in the dorsal spinal cord after peripheral nerve injury cross-sex and cross-species, and imputed commonalities, as well as differences in cellular pathways and gene regulation. KEY RESULTS: We identified 93 transcripts in the dorsal horn that were increased by peripheral nerve injury in male and female mice and rats. Following gene ontology and transcription factor analyses, we constructed a pain interactome for the proteins encoded by the differentially expressed genes, discovering new, conserved signalling nodes. We investigated the interactome with the Drug-Gene database to predict FDA-approved medications that may modulate key nodes within the network. The top hit from the analysis was fostamatinib, the molecular target of which is the non-receptor spleen associated tyrosine kinase (Syk), which our analysis had identified as a key node in the interactome. We found that intrathecally administrating the active metabolite of fostamatinib, R406 and another Syk inhibitor P505-15, significantly reversed pain hypersensitivity in both sexes. CONCLUSIONS AND IMPLICATIONS: Thus, we have identified and shown the efficacy of an agent that could not have been previously predicted to have analgesic properties.

The promises and challenges of clinical AI in community paediatric medicine.

The widespread adoption of virtual care technologies has quickly reshaped healthcare operations and delivery, particularly in the context of community medicine. In this paper, we use the virtual care landscape as a point of departure to envision the promises and challenges of artificial intelligence (AI) in healthcare. Our analysis is directed towards community care practitioners interested in learning more about how AI can change their practice along with the critical considerations required to integrate AI into their practice. We highlight examples of how AI can enable access to new sources of clinical data while augmenting clinical workflows and healthcare delivery. AI can help optimize how and when care is delivered by community practitioners while also improving practice efficiency, accessibility, and the overall quality of care. Unlike virtual care, however, AI is still missing many of the key enablers required to facilitate adoption into the community care landscape and there are challenges we must consider and resolve for AI to successfully improve healthcare delivery. We discuss several critical considerations, including data governance in the clinic setting, healthcare practitioner education, regulation of AI in healthcare, clinician reimbursement, and access to both technology and the internet.

Developing OCHROdb, a comprehensive quality checked database of open chromatin regions from sequencing data.

International consortia, including ENCODE, Roadmap Epigenomics, Genomics of Gene Regulation and Blueprint Epigenome have made large-scale datasets of open chromatin regions publicly available. While these datasets are extremely useful for studying mechanisms of gene regulation in disease and cell development, they only identify open chromatin regions in individual samples. A uniform comparison of accessibility of the same regulatory sites across multiple samples is necessary to correlate open chromatin accessibility and expression of target genes across matched cell types. Additionally, although replicate samples are available for majority of cell types, a comprehensive replication-based quality checking of individual regulatory sites is still lacking. We have integrated 828 DNase-I hypersensitive sequencing samples, which we have uniformly processed and then clustered their regulatory regions across all samples. We checked the quality of open-chromatin regions using our replication test. This has resulted in a comprehensive, quality-checked database of Open CHROmatin (OCHROdb) regions for 194 unique human cell types and cell lines which can serve as a reference for gene regulatory studies involving open chromatin. We have made this resource publicly available: users can download the whole database, or query it for their genomic regions of interest and visualize the results in an interactive genome browser.

COVID-19 vaccine antibody responses in community-dwelling adults to 48 weeks post primary vaccine series.

We report a decentralized prospective cohort study of self-reported adverse events and antibody responses to COVID vaccines derived from dried blood spots. Data are presented for 911 older (aged >70 years) and 375 younger (30-50 years) recruits to 48 weeks after the primary vaccine series. After a single vaccine, 83% younger and 45% older participants had overall seropositivity (p < 0.0001) increasing to 100/98% with the second dose, respectively (p = 0.084). A cancer diagnosis (p = 0.009), no mRNA-1273 vaccine doses (p <0 .0001), and older age (p <0 .0001) predicted lower responses. Antibody levels declined in both cohorts at 12 and 24 weeks increasing with booster doses. At 48 weeks, for participants with 3 vaccine doses, the median antibody levels were higher in the older cohort (p = 0.04) with any dose of mRNA-1273 (p <0 .0001) and with COVID infection (p <0 .001). The vaccines were well tolerated. Breakthrough COVID infections were uncommon (16% older cohort, 29% younger cohort; p < 0.0001) and mild.

Trio RNA sequencing in a cohort of medically complex children.

Genome sequencing (GS) is a powerful test for the diagnosis of rare genetic disorders. Although GS can enumerate most non-coding variation, determining which non-coding variants are disease-causing is challenging. RNA sequencing (RNA-seq) has emerged as an important tool to help address this issue, but its diagnostic utility remains understudied, and the added value of a trio design is unknown. We performed GS plus RNA-seq from blood using an automated clinical-grade high-throughput platform on 97 individuals from 39 families where the proband was a child with unexplained medical complexity. RNA-seq was an effective adjunct test when paired with GS. It enabled clarification of putative splice variants in three families, but it did not reveal variants not already identified by GS analysis. Trio RNA-seq decreased the number of candidates requiring manual review when filtering for de novo dominant disease-causing variants, allowing for the exclusion of 16% of gene-expression outliers and 27% of allele-specific-expression outliers. However, clear diagnostic benefit from the trio design was not observed. Blood-based RNA-seq can facilitate genome analysis in children with suspected undiagnosed genetic disease. In contrast to DNA sequencing, the clinical advantages of a trio RNA-seq design may be more limited.

Evaluation of single-cell RNAseq labelling algorithms using cancer datasets.

Single-cell RNA sequencing (scRNA-seq) clustering and labelling methods are used to determine precise cellular composition of tissue samples. Automated labelling methods rely on either unsupervised, cluster-based approaches or supervised, cell-based approaches to identify cell types. The high complexity of cancer poses a unique challenge, as tumor microenvironments are often composed of diverse cell subpopulations with unique functional effects that may lead to disease progression, metastasis and treatment resistance. Here, we assess 17 cell-based and 9 cluster-based scRNA-seq labelling algorithms using 8 cancer datasets, providing a comprehensive large-scale assessment of such methods in a cancer-specific context. Using several performance metrics, we show that cell-based methods generally achieved higher performance and were faster compared to cluster-based methods. Cluster-based methods more successfully labelled non-malignant cell types, likely because of a lack of gene signatures for relevant malignant cell subpopulations. Larger cell numbers present in some cell types in training data positively impacted prediction scores for cell-based methods. Finally, we examined which methods performed favorably when trained and tested on separate patient cohorts in scenarios similar to clinical applications, and which were able to accurately label particularly small or under-represented cell populations in the given datasets. We conclude that scPred and SVM show the best overall performances with cancer-specific data and provide further suggestions for algorithm selection. Our analysis pipeline for assessing the performance of cell type labelling algorithms is available in https://github.com/shooshtarilab/scRNAseq-Automated-Cell-Type-Labelling.

ChartWalk: Navigating large collections of text notes in electronic health records for clinical chart review.

Before seeing a patient for the first time, healthcare workers will typically conduct a comprehensive clinical chart review of the patient's electronic health record (EHR). Within the diverse documentation pieces included there, text notes are among the most important and thoroughly perused segments for this task; and yet they are among the least supported medium in terms of content navigation and overview. In this work, we delve deeper into the task of clinical chart review from a data visualization perspective and propose a hybrid graphics+text approach via ChartWalk, an interactive tool to support the review of text notes in EHRs. We report on our iterative design process grounded in input provided by a diverse range of healthcare professionals, with steps including: (a) initial requirements distilled from interviews and the literature, (b) an interim evaluation to validate design decisions, and (c) a task-based qualitative evaluation of our final design. We contribute lessons learned to better support the design of tools not only for clinical chart reviews but also other healthcare-related tasks around medical text analysis.

Care4Rare Canada: Outcomes from a decade of network science for rare disease gene discovery.

The past decade has witnessed a rapid evolution in rare disease (RD) research, fueled by the availability of genome-wide (exome and genome) sequencing. In 2011, as this transformative technology was introduced to the research community, the Care4Rare Canada Consortium was launched: initially as FORGE, followed by Care4Rare, and Care4Rare SOLVE. Over what amounted to three eras of diagnosis and discovery, the Care4Rare Consortium used exome sequencing and, more recently, genome and other 'omic technologies to identify the molecular cause of unsolved RDs. We achieved a diagnostic yield of 34% (623/1,806 of participating families), including the discovery of deleterious variants in 121 genes not previously associated with disease, and we continue to study candidate variants in novel genes for 145 families. The Consortium has made significant contributions to RD research, including development of platforms for data collection and sharing and instigating a Canadian network to catalyze functional characterization research of novel genes. The Consortium was instrumental to implementing genome-wide sequencing as a publicly funded test for RD diagnosis in Canada. Despite the successes of the past decade, the challenge of solving all RDs remains enormous, and the work is far from over. We must leverage clinical and 'omic data for secondary use, develop tools and policies to support safe data sharing, continue to explore the utility of new and emerging technologies, and optimize research protocols to delineate complex disease mechanisms. Successful approaches in each of these realms is required to offer diagnostic clarity to all families with RDs.

Evaluation of single-cell RNA-seq clustering algorithms on cancer tumor datasets.

Tumors are complex biological entities that comprise cell types of different origins, with different mutational profiles and different patterns of transcriptional dysregulation. The exploration of data related to cancer biology requires careful analytical methods to reflect the heterogeneity of cell populations in cancer samples. Single-cell techniques are now able to capture the transcriptional profiles of individual cells. However, the complexity of RNA-seq data, especially in cancer samples, makes it challenging to cluster single-cell profiles into groups that reflect the underlying cell types. We have developed a framework for a systematic examination of single-cell RNA-seq clustering algorithms for cancer data, which uses a range of well-established metrics to generate a unified quality score and algorithm ranking. To demonstrate this framework, we examined clustering performance of 15 different single-cell RNA-seq clustering algorithms on eight different cancer datasets. Our results suggest that the single-cell RNA-seq clustering algorithms fall into distinct groups by performance, with the highest clustering quality on non-malignant cells achieved by three algorithms: Seurat, bigSCale and Cell Ranger. However, for malignant cells, two additional algorithms often reach a better performance, namely Monocle and SC3. Their ability to detect known rare cell types was also among the best, along with Seurat. Our approach and results can be used by a broad audience of practitioners who analyze single-cell transcriptomic data in cancer research.

Assessment of Machine Learning-Based Medical Directives to Expedite Care in Pediatric Emergency Medicine.

Importance: Increased wait times and long lengths of stay in emergency departments (EDs) are associated with poor patient outcomes. Systems to improve ED efficiency would be useful. Specifically, minimizing the time to diagnosis by developing novel workflows that expedite test ordering can help accelerate clinical decision-making. Objective: To explore the use of machine learning-based medical directives (MLMDs) to automate diagnostic testing at triage for patients with common pediatric ED diagnoses. Design, Setting, and Participants: Machine learning models trained on retrospective electronic health record data were evaluated in a decision analytical model study conducted at the ED of the Hospital for Sick Children Toronto, Canada. Data were collected on all patients aged 0 to 18 years presenting to the ED from July 1, 2018, to June 30, 2019 (77 219 total patient visits). Exposure: Machine learning models were trained to predict the need for urinary dipstick testing, electrocardiogram, abdominal ultrasonography, testicular ultrasonography, bilirubin level testing, and forearm radiographs. Main Outcomes and Measures: Models were evaluated using area under the receiver operator curve, true-positive rate, false-positive rate, and positive predictive values. Model decision thresholds were determined to limit the total number of false-positive results and achieve high positive predictive values. The time difference between patient triage completion and test ordering was assessed for each use of MLMD. Error rates were analyzed to assess model bias. In addition, model explainability was determined using Shapley Additive Explanations values. Results: There was a total of 42 238 boys (54.7%) included in model development; mean (SD) age of the children was 5.4 (4.8) years. Models obtained high area under the receiver operator curve (0.89-0.99) and positive predictive values (0.77-0.94) across each of the use cases. The proposed implementation of MLMDs would streamline care for 22.3% of all patient visits and make test results available earlier by 165 minutes (weighted mean) per affected patient. Model explainability for each MLMD demonstrated clinically relevant features having the most influence on model predictions. Models also performed with minimal to no sex bias. Conclusions and Relevance: The findings of this study suggest the potential for clinical automation using MLMDs. When integrated into clinical workflows, MLMDs may have the potential to autonomously order common ED tests early in a patient's visit with explainability provided to patients and clinicians.

Splicing is an alternate oncogenic pathway activation mechanism in glioma.

High-grade diffuse glioma (HGG) is the leading cause of brain tumour death. While the genetic drivers of HGG have been well described, targeting these has thus far had little impact on survival suggesting other mechanisms are at play. Here we interrogate the alternative splicing landscape of pediatric and adult HGG through multi-omic analyses, uncovering an increased splicing burden compared with normal brain. The rate of recurrent alternative splicing in cancer drivers exceeds their mutation rate, a pattern that is recapitulated in pan-cancer analyses, and is associated with worse prognosis in HGG. We investigate potential oncogenicity by interrogating cancer pathways affected by alternative splicing in HGG; spliced cancer drivers include members of the RAS/MAPK pathway. RAS suppressor neurofibromin 1 is differentially spliced to a less active isoform in >80% of HGG downstream from REST upregulation, activating the RAS/MAPK pathway and reducing glioblastoma patient survival. Overall, our results identify non-mutagenic mechanisms by which cancers activate oncogenic pathways which need to accounted for in personalized medicine approaches.

PhenomeCentral: 7 years of rare disease matchmaking.

A major challenge in validating genetic causes for patients with rare diseases (RDs) is the difficulty in identifying other RD patients with overlapping phenotypes and variants in the same candidate gene. This process, known as matchmaking, requires robust data sharing solutions to be effective. In 2014 we launched PhenomeCentral, a RD data repository capable of collecting computer-readable genotypic and phenotypic data for the purposes of RD matchmaking. Over the past 7 years PhenomeCentral's features have been expanded and its data set has consistently grown. There are currently 1615 users registered on PhenomeCentral, which have contributed over 12,000 patient cases. Most of these cases contain detailed phenotypic terms, with a significant portion also providing genomic sequence data or other forms of clinical information. Matchmaking within PhenomeCentral, and with connections to other data repositories in the Matchmaker Exchange, have collectively resulted in over 60,000 matches, which have facilitated multiple gene discoveries. The collection of deep phenotypic and genotypic data has also positioned PhenomeCentral well to support next generation of matchmaking initiatives that utilize genome sequencing data, ensuring that PhenomeCentral will remain a useful tool in solving undiagnosed RD cases in the years to come.

Genomics4RD: An integrated platform to share Canadian deep-phenotype and multiomic data for international rare disease gene discovery.

Despite recent progress in the understanding of the genetic etiologies of rare diseases (RDs), a significant number remain intractable to diagnostic and discovery efforts. Broad data collection and sharing of information among RD researchers is therefore critical. In 2018, the Care4Rare Canada Consortium launched the project C4R-SOLVE, a subaim of which was to collect, harmonize, and share both retrospective and prospective Canadian clinical and multiomic data. Here, we introduce Genomics4RD, an integrated web-accessible platform to share Canadian phenotypic and multiomic data between researchers, both within Canada and internationally, for the purpose of discovering the mechanisms that cause RDs. Genomics4RD has been designed to standardize data collection and processing, and to help users systematically collect, prioritize, and visualize participant information. Data storage, authorization, and access procedures have been developed in collaboration with policy experts and stakeholders to ensure the trusted and secure access of data by external researchers. The breadth and standardization of data offered by Genomics4RD allows researchers to compare candidate disease genes and variants between participants (i.e., matchmaking) for discovery purposes, while facilitating the development of computational approaches for multiomic data analyses and enabling clinical translation efforts for new genetic technologies in the future.

PhenoPad: Building AI enabled note-taking interfaces for patient encounters.

Current clinical note-taking approaches cannot capture the entirety of information available from patient encounters and detract from patient-clinician interactions. By surveying healthcare providers' current note-taking practices and attitudes toward new clinical technologies, we developed a patient-centered paradigm for clinical note-taking that makes use of hybrid tablet/keyboard devices and artificial intelligence (AI) technologies. PhenoPad is an intelligent clinical note-taking interface that captures free-form notes and standard phenotypic information via a variety of modalities, including speech and natural language processing techniques, handwriting recognition, and more. The output is unobtrusively presented on mobile devices to clinicians for real-time validation and can be automatically transformed into digital formats that would be compatible with integration into electronic health record systems. Semi-structured interviews and trials in clinical settings rendered positive feedback from both clinicians and patients, demonstrating that AI-enabled clinical note-taking under our design improves ease and breadth of information captured during clinical visits without compromising patient-clinician interactions. We open source a proof-of-concept implementation that can lay the foundation for broader clinical use cases.

Outcome of over 1500 matches through the Matchmaker Exchange for rare disease gene discovery: The 2-year experience of Care4Rare Canada.

PURPOSE: Matchmaking has emerged as a useful strategy for building evidence toward causality of novel disease genes in patients with undiagnosed rare diseases. The Matchmaker Exchange (MME) is a collaborative initiative that facilitates international data sharing for matchmaking purposes; however, data on user experience is limited. METHODS: Patients enrolled as part of the Finding of Rare Disease Genes in Canada (FORGE) and Care4Rare Canada research programs had their exome sequencing data reanalyzed by a multidisciplinary research team over a 2-year period. Compelling variants in genes not previously associated with a human phenotype were submitted through the MME node PhenomeCentral, and outcomes were collected. RESULTS: In this study, 194 novel candidate genes were submitted to the MME, resulting in 1514 matches, and 15% of the genes submitted resulted in collaborations. Most submissions resulted in at least 1 match, and most matches were with GeneMatcher (82%), where additional email exchange was required to evaluate the match because of the lack of phenotypic or inheritance information. CONCLUSION: Matchmaking through the MME is an effective way to investigate novel candidate genes; however, it is a labor-intensive process. Engagement from the community to contribute phenotypic, genotypic, and inheritance data will ensure that matchmaking continues to be a useful approach in the future.