Returning Individual Research Results from Digital Phenotyping in Psychiatry.

Psychiatry is rapidly adopting digital phenotyping and artificial intelligence/machine learning tools to study mental illness based on tracking participants' locations, online activity, phone and text message usage, heart rate, sleep, physical activity, and more. Existing ethical frameworks for return of individual research results (IRRs) are inadequate to guide researchers for when, if, and how to return this unprecedented number of potentially sensitive results about each participant's real-world behavior. To address this gap, we convened an interdisciplinary expert working group, supported by a National Institute of Mental Health grant. Building on established guidelines and the emerging norm of returning results in participant-centered research, we present a novel framework specific to the ethical, legal, and social implications of returning IRRs in digital phenotyping research. Our framework offers researchers, clinicians, and Institutional Review Boards (IRBs) urgently needed guidance, and the principles developed here in the context of psychiatry will be readily adaptable to other therapeutic areas.

High number of candidate gene variants are identified as disease-causing in a period of 4 years.

Advances in bioinformatic tools paired with the ongoing accumulation of genetic knowledge and periodic reanalysis of genomic sequencing data have led to an improvement in genetic diagnostic rates. Candidate gene variants (CGVs) identified during sequencing or on reanalysis but not yet implicated in human disease or associated with a phenotypically distinct condition are often not revisited, leading to missed diagnostic opportunities. Here, we revisited 33 such CGVs from our previously published study and determined that 16 of them are indeed disease-causing (novel or phenotype expansion) since their identification. These results emphasize the need to focus on previously identified CGVs during sequencing or reanalysis and the importance of sharing that information with researchers around the world, including relevant functional analysis to establish disease causality.

Clinical Characterization of Pediatric Erythromelalgia: A Single-Center Case Series.

Erythromelalgia is a descriptive term for severe burning pain and erythema in the distal extremities relieved by cold and exacerbated by heat. Pediatric case series to date are relatively small. We extracted and analyzed medical record data for 42 pediatric patients to describe clinical characteristics, associated conditions, and responses to treatments. Informed consent was obtained according to an IRB-approved protocol that included gene discovery. Three patients had confirmed Nav1.7 sodium channelopathies, with six additional patients under investigation with novel gene candidates. There was a female predominance (2.5:1), and the median onset age was 12 years (IQR = 3-14). Patients saw a median of three specialists (IQR = 2-3) for a diagnosis. The majority (90%) reported bilateral symptoms. Cooling methods usually provided partial relief, while heat and exercise exacerbated pain. No medication appeared to be consistently effective; commonly prescribed medications included sodium channel blockers (n = 37), topical analgesics (n = 26), gabapentin (n = 22), and aspirin (n = 15). Based on the currently published literature, we believe this cohort is the largest pediatric study of erythromelalgia to date. Many findings are consistent with those of previously published case series. Work is in progress to establish a prospective cohort and multi-center registry.

Hemizygous variants in protein phosphatase 1 regulatory subunit 3F (PPP1R3F) are associated with a neurodevelopmental disorder characterized by developmental delay, intellectual disability and autistic features.

Protein phosphatase 1 regulatory subunit 3F (PPP1R3F) is a member of the glycogen targeting subunits (GTSs), which belong to the large group of regulatory subunits of protein phosphatase 1 (PP1), a major eukaryotic serine/threonine protein phosphatase that regulates diverse cellular processes. Here, we describe the identification of hemizygous variants in PPP1R3F associated with a novel X-linked recessive neurodevelopmental disorder in 13 unrelated individuals. This disorder is characterized by developmental delay, mild intellectual disability, neurobehavioral issues such as autism spectrum disorder, seizures and other neurological findings including tone, gait and cerebellar abnormalities. PPP1R3F variants segregated with disease in affected hemizygous males that inherited the variants from their heterozygous carrier mothers. We show that PPP1R3F is predominantly expressed in brain astrocytes and localizes to the endoplasmic reticulum in cells. Glycogen content in PPP1R3F knockout astrocytoma cells appears to be more sensitive to fluxes in extracellular glucose levels than in wild-type cells, suggesting that PPP1R3F functions in maintaining steady brain glycogen levels under changing glucose conditions. We performed functional studies on nine of the identified variants and observed defects in PP1 binding, protein stability, subcellular localization and regulation of glycogen metabolism in most of them. Collectively, the genetic and molecular data indicate that deleterious variants in PPP1R3F are associated with a new X-linked disorder of glycogen metabolism, highlighting the critical role of GTSs in neurological development. This research expands our understanding of neurodevelopmental disorders and the role of PP1 in brain development and proper function.

Children with Early-Onset Psychosis Have Increased Burden of Rare GRIN2A Variants.

BACKGROUND: Children and adolescents with early-onset psychosis (EOP) have more rare genetic variants than individuals with adult-onset forms of the illness, implying that fewer EOP participants are needed for genetic discovery. The Schizophrenia Exome Sequencing Meta-analysis (SCHEMA) study predicted that 10 genes with ultra-rare variation were linked to adult-onset schizophrenia. We hypothesized that rare variants predicted "High" and "Moderate" by the Variant Effect Predictor Algorithm (abbreviated as VEPHMI) in these 10 genes would be enriched in our EOP cohort. METHODS: We compared rare VEPHMI variants in individuals with EOP (N = 34) with race- and sex-matched controls (N = 34) using the sequence kernel association test (SKAT). RESULTS: GRIN2A variants were significantly increased in the EOP cohort (p = 0.004), with seven individuals (20% of the EOP cohort) carrying a rare VEPHMI variant. The EOP cohort was then compared to three additional control cohorts. GRIN2A variants were significantly increased in the EOP cohort for two of the additional control sets (p = 0.02 and p = 0.02), and trending towards significance for the third (p = 0.06). CONCLUSION: Despite a small sample size, GRIN2A VEPHMI variant burden was increased in a cohort of individuals with EOP in comparison to controls. GRIN2A variants have been associated with a range of neuropsychiatric disorders including adult-onset psychotic spectrum disorder and childhood-onset schizophrenia. This study supports the role of GRIN2A in EOP and emphasizes its role in neuropsychiatric disorders.

X-linked hypophosphatemia in 4 generations due to an exon 13-15 duplication in PHEX, in the absence of the c.*231A>G variant.

X-linked hypophosphatemia is the most common cause of inherited rickets, due to inactivating variants of PHEX. More than 800 variants have been described to date and one which consists of a single base change in the 3' untranslated region (UTR) (c.*231A>G) is reported as prevalent in North America. Recently an exon 13-15 duplication has been found to occur in concert with the c.*231A>G variant, and thus it is unclear whether the pathogenicity is solely a function of the UTR variant. We present a family with XLH who harbors the exon 13-15 duplication but does not carry the 3'UTR variant, providing evidence that the duplication itself is the pathogenic variant when these two variants are found in cis.

Mendelian Disorders in an Interstitial Cystitis/Bladder Pain Syndrome Cohort.

Interstitial cystitis/bladder pain syndrome (IC/BPS) is a chronic pain disorder causing symptoms of urinary frequency, urgency, and bladder discomfort or pain. Although this condition affects a large population, little is known about its etiology. Genetic analyses of whole exome sequencing are performed on 109 individuals with IC/BPS. One family has a previously reported SIX5 variant (ENST00000317578.6:c.472G>A, p.Ala158Thr), consistent with Branchiootorenal syndrome 2 (BOR2). A likely pathogenic heterozygous variant in ATP2A2 (ENST00000539276.2:c.235G>A, p.Glu79Lys) is identified in two unrelated probands, indicating possible Darier-White disease. Two private heterozygous variants are identified in ATP2C1 (ENST00000393221.4:c.2358A>T, p.Glu786Asp (VUS/Likely Pathogenic) and ENST00000393221.4:c.989C>G, p.Thr330Ser (likely pathogenic)), indicative of Hailey-Hailey Disease. Sequence kernel association test analysis finds an increased burden of rare ATP2C1 variants in the IC/BPS cases versus a control cohort (p = 0.03, OR = 6.76), though does not survive Bonferroni correction. The data suggest that some individuals with IC/BPS may have unrecognized Mendelian syndromes. Comprehensive phenotyping and genotyping aid in understanding the range of diagnoses in the population-based IC/BPS cohort. Conversely, ATP2C1, ATP2A2, and SIX5 may be candidate genes for IC/BPS. Further evaluation with larger numbers is needed. Genetically screening individuals with IC/BPS may help diagnose and treat this painful disorder due to its heterogeneous nature.

Copy Number Variation and Structural Genomic Findings in 116 Cases of Sudden Unexplained Death between 1 and 28 Months of Age.

In sudden unexplained death in pediatrics (SUDP) the cause of death is unknown despite an autopsy and investigation. The role of copy number variations (CNVs) in SUDP has not been well-studied. Chromosomal microarray (CMA) data are generated for 116 SUDP cases with age at death between 1 and 28 months. CNVs are classified using the American College of Medical Genetics and Genomics guidelines and CNVs in our cohort are compared to an autism spectrum disorder (ASD) cohort, and to a control cohort. Pathogenic CNVs are identified in 5 of 116 cases (4.3%). Variants of uncertain significance (VUS) favoring pathogenic CNVs are identified in 9 cases (7.8%). Several CNVs are associated with neurodevelopmental phenotypes including seizures, ASD, developmental delay, and schizophrenia. The structural variant 47,XXY is identified in two cases (2/69 boys, 2.9%) not previously diagnosed with Klinefelter syndrome. Pathogenicity scores for deletions are significantly elevated in the SUDP cohort versus controls (p = 0.007) and are not significantly different from the ASD cohort. The finding of pathogenic or VUS favoring pathogenic CNVs, or structural variants, in 12.1% of cases, combined with the observation of higher pathogenicity scores for deletions in SUDP versus controls, suggests that CMA should be included in the genetic evaluation of SUDP.

Genetic Predisposition to Neurological Complications in Patients with COVID-19.

Several studies have identified rare and common genetic variants associated with severe COVID-19, but no study has reported genetic determinants as predisposition factors for neurological complications. In this report, we identified rare/unique structural variants (SVs) implicated in neurological functions in two individuals with neurological manifestations of COVID-19. This report highlights the possible genetic link to the neurological symptoms with COVID-19 and calls for a collective effort to study these cohorts for a possible genetic linkage.

Reanalysis of clinical exome identifies the second variant in two individuals with recessive disorders.

Clinical exome/genome sequencing is increasingly being utilized by clinicians to diagnose various likely genetic conditions, but many cases remain undiagnosed. In a subset of those undiagnosed cases, a single heterozygous variant in an autosomal recessive (AR) condition with consistent phenotype may be identified, raising the question if a second variant is missing. Here, we report two cases of recessive conditions in which only one heterozygous variant was initially reported by clinical exome sequencing, and on research reanalysis a second heterozygous variant in trans was identified. We performed a review of the existing exome reanalysis literature and found that this aspect is often not emphasized. These findings highlight the importance of data reanalysis in undiagnosed cases where only a single disease-associated variant is identified in an AR condition with a strong link to presenting phenotype.

16p13.11 deletion variants associated with neuropsychiatric disorders cause morphological and synaptic changes in induced pluripotent stem cell-derived neurons.

16p13.11 copy number variants (CNVs) have been associated with autism, schizophrenia, psychosis, intellectual disability, and epilepsy. The majority of 16p13.11 deletions or duplications occur within three well-defined intervals, and despite growing knowledge of the functions of individual genes within these intervals, the molecular mechanisms that underlie commonly observed clinical phenotypes remain largely unknown. Patient-derived, induced pluripotent stem cells (iPSCs) provide a platform for investigating the morphological, electrophysiological, and gene-expression changes that result from 16p13.11 CNVs in human-derived neurons. Patient derived iPSCs with varying sizes of 16p13.11 deletions and familial controls were differentiated into cortical neurons for phenotypic analysis. High-content imaging and morphological analysis of patient-derived neurons demonstrated an increase in neurite branching in patients compared with controls. Whole-transcriptome sequencing revealed expression level changes in neuron development and synaptic-related gene families, suggesting a defect in synapse formation. Subsequent quantification of synapse number demonstrated increased numbers of synapses on neurons derived from early-onset patients compared to controls. The identification of common phenotypes among neurons derived from patients with overlapping 16p13.11 deletions will further assist in ascertaining common pathways and targets that could be utilized for screening drug candidates. These studies can help to improve future treatment options and clinical outcomes for 16p13.11 deletion patients.

The Phenotypic Continuum of ATP1A3-Related Disorders.

BACKGROUND AND OBJECTIVES: ATP1A3 is associated with a broad spectrum of predominantly neurologic disorders, which continues to expand beyond the initially defined phenotypes of alternating hemiplegia of childhood, rapid-onset dystonia parkinsonism, and cerebellar ataxia, areflexia, pes cavus, optic atrophy, sensorineural hearing loss syndrome. This phenotypic variability makes it challenging to assess the pathogenicity of an ATP1A3 variant found in an undiagnosed patient. We describe the phenotypic features of individuals carrying a pathogenic/likely pathogenic ATP1A3 variant and perform a literature review of all ATP1A3 variants published thus far in association with human neurologic disease. Our aim is to demonstrate the heterogeneous clinical spectrum of the gene and look for phenotypic overlap between patients that will streamline the diagnostic process. METHODS: Undiagnosed individuals with ATP1A3 variants were identified within the cohort of the Deciphering Developmental Disorders study with additional cases contributed by collaborators internationally. Detailed clinical data were collected with consent through a questionnaire completed by the referring clinicians. PubMed was searched for publications containing the term "ATP1A3" from 2004 to 2021. RESULTS: Twenty-four individuals with a previously undiagnosed neurologic phenotype were found to carry 21 ATP1A3 variants. Eight variants have been previously published. Patients experienced on average 2-3 different types of paroxysmal events. Permanent neurologic features were common including microcephaly (7; 29%), ataxia (13; 54%), dystonia (10; 42%), and hypotonia (7; 29%). All patients had cognitive impairment. Neuropsychiatric diagnoses were reported in 16 (66.6%) individuals. Phenotypes were extremely varied, and most individuals did not fit clinical criteria for previously published phenotypes. On review of the literature, 1,108 individuals have been reported carrying 168 different ATP1A3 variants. The most common variants are associated with well-defined phenotypes, while more rare variants often result in very rare symptom correlations, such as are seen in our study. Combined Annotation-Dependent Depletion (CADD) scores of pathogenic and likely pathogenic variants were significantly higher and variants clustered within 6 regions of constraint. DISCUSSION: Our study shows that looking for a combination of paroxysmal events, hyperkinesia, neuropsychiatric symptoms, and cognitive impairment and evaluating the CADD score and variant location can help identify an ATP1A3-related condition, rather than applying diagnostic criteria alone.

Similar Rates of Deleterious Copy Number Variants in Early-Onset Psychosis and Autism Spectrum Disorder.

OBJECTIVE: Copy number variants (CNVs) are strongly associated with neurodevelopmental and psychotic disorders. Early-onset psychosis (EOP), where symptoms appear before 18 years of age, is thought to be more strongly influenced by genetic factors than adult-onset psychotic disorders. However, the prevalence and effect of CNVs in EOP is unclear. METHODS: The authors documented the prevalence of recurrent CNVs and the functional impact of deletions and duplications genome-wide in 137 children and adolescents with EOP compared with 5,540 individuals with autism spectrum disorder (ASD) and 16,504 population control subjects. Specifically, the frequency of 47 recurrent CNVs previously associated with neurodevelopmental and neuropsychiatric illnesses in each cohort were compared. Next, CNV risk scores (CRSs), indices reflecting the dosage sensitivity for any gene across the genome that is encapsulated in a deletion or duplication separately, were compared between groups. RESULTS: The prevalence of recurrent CNVs was significantly higher in the EOP group than in the ASD (odds ratio=2.30) and control (odds ratio=5.06) groups. However, the difference between the EOP and ASD groups was attenuated when EOP participants with co-occurring ASD were excluded. CRS was significantly higher in the EOP group compared with the control group for both deletions (odds ratio=1.30) and duplications (odds ratio=1.09). In contrast, the EOP and ASD groups did not differ significantly in terms of CRS. CONCLUSIONS: Given the high frequency of recurrent CNVs in the EOP group and comparable CRSs in the EOP and ASD groups, the findings suggest that all children and adolescents with a psychotic diagnosis should undergo genetic screening, as is recommended in ASD.

Drosophila functional screening of de novo variants in autism uncovers damaging variants and facilitates discovery of rare neurodevelopmental diseases.

Individuals with autism spectrum disorder (ASD) exhibit an increased burden of de novo mutations (DNMs) in a broadening range of genes. While these studies have implicated hundreds of genes in ASD pathogenesis, which DNMs cause functional consequences in vivo remains unclear. We functionally test the effects of ASD missense DNMs using Drosophila through "humanization" rescue and overexpression-based strategies. We examine 79 ASD variants in 74 genes identified in the Simons Simplex Collection and find 38% of them to cause functional alterations. Moreover, we identify GLRA2 as the cause of a spectrum of neurodevelopmental phenotypes beyond ASD in 13 previously undiagnosed subjects. Functional characterization of variants in ASD candidate genes points to conserved neurobiological mechanisms and facilitates gene discovery for rare neurodevelopmental diseases.

Novel CAPN1 missense variants in complex hereditary spastic paraplegia with early-onset psychosis.

CAPN1-associated hereditary spastic paraplegia (SPG76) is a rare and clinically heterogenous syndrome due to loss of calpain-1 function. Here we illustrate a translational approach to the case of an 18-year-old patient who first presented with psychiatric symptoms followed by spastic gait, intention tremor, and neurogenic bladder dysfunction, consistent with a complex form of HSP. Exome sequencing showed compound-heterozygous missense variants in CAPN1 (NM_001198868.2: c.1712A>G (p.Asn571Ser)/c.1991C>T (p.Ser664Leu)) and a previously reported heterozygous stop-gain variant in RCL1. In silico analyses of the CAPN1 variants predicted a deleterious effect and in vitro functional studies confirmed reduced calpain-1 activity and dysregulated downstream signaling. These findings support a diagnosis of SPG76 and highlight that the psychiatric symptoms can precede the motor symptoms in HSP. Our results also suggest that multiple genes can potentially contribute to complex neuropsychiatric diseases.

Genetic Determinants of Sudden Unexpected Death in Pediatrics.

PURPOSE: This study aimed to evaluate genetic contributions to sudden unexpected death in pediatrics (SUDP). METHODS: We phenotyped and performed exome sequencing for 352 SUDP cases. We analyzed variants in 294 "SUDP genes" with mechanisms plausibly related to sudden death. In a subset of 73 cases with parental data (trios), we performed exome-wide analyses and conducted cohort-wide burden analyses. RESULTS: In total, we identified likely contributory variants in 37 of 352 probands (11%). Analysis of SUDP genes identified pathogenic/likely pathogenic variants in 12 of 352 cases (SCN1A, DEPDC5 [2], GABRG2, SCN5A [2], TTN [2], MYBPC3, PLN, TNNI3, and PDHA1) and variants of unknown significance-favor-pathogenic in 17 of 352 cases. Exome-wide analyses of the 73 cases with family data additionally identified 4 de novo pathogenic/likely pathogenic variants (SCN1A [2], ANKRD1, and BRPF1) and 4 de novo variants of unknown significance-favor-pathogenic. Comparing cases with controls, we demonstrated an excess burden of rare damaging SUDP gene variants (odds ratio, 2.94; 95% confidence interval, 2.37-4.21) and of exome-wide de novo variants in the subset of 73 with trio data (odds ratio, 3.13; 95% confidence interval, 1.91-5.16). CONCLUSION: We provide strong evidence for a role of genetic factors in SUDP, involving both candidate genes and novel genes for SUDP and expanding phenotypes of disease genes not previously associated with sudden death.

Optical genome mapping identifies rare structural variations as predisposition factors associated with severe COVID-19.

Impressive global efforts have identified both rare and common gene variants associated with severe COVID-19 using sequencing technologies. However, these studies lack the sensitivity to accurately detect several classes of variants, especially large structural variants (SVs), which account for a substantial proportion of genetic diversity including clinically relevant variation. We performed optical genome mapping on 52 severely ill COVID-19 patients to identify rare/unique SVs as decisive predisposition factors associated with COVID-19. We identified 7 SVs involving genes implicated in two key host-viral interaction pathways: innate immunity and inflammatory response, and viral replication and spread in nine patients, of which SVs in STK26 and DPP4 genes are the most intriguing candidates. This study is the first to systematically assess the potential role of SVs in the pathogenesis of COVID-19 severity and highlights the need to evaluate SVs along with sequencing variants to comprehensively associate genomic information with interindividual variability in COVID-19 phenotypes.

ITSN1: a novel candidate gene involved in autosomal dominant neurodevelopmental disorder spectrum.

ITSN1 plays an important role in brain development. Recent studies in large cohorts of subjects with neurodevelopmental disorders have identified de novo variants in ITSN1 gene thereby suggesting that this gene is involved in the development of such disorders. The aim of this study is to provide further proof of such a link. We performed trio exome sequencing in a patient presenting autism, intellectual disability, and severe behavioral difficulties. Additional affected patients with a neurodevelopmental disorder harboring a heterozygous variant in ITSN1 (NM_003024.2) were collected through a worldwide collaboration. All patients underwent detailed phenotypic and genetic assessment and data was collected and shared by healthcare givers. We identified ten novel patients from eight families with heterozygous truncating or missense variants in ITSN1 gene. In addition, four previously published patients from large meta-analysis studies were included. In total, 7/14 patients presented a de novo variant in ITSN1. All patients showed neurodevelopmental disorders from autism spectrum disorders (90%), intellectual disability (86%), and epilepsy (30%). We demonstrated that truncating variants are in the first half of ITSN1 whereas missense variants are clustered in C-terminal region. We suggest ITSN1 gene is involved in development of an autism spectrum disorder with variable additional neurodevelopmental deficiency, thus confirming the hypothesis that ITSN1 is important for brain development.

Artificial intelligence enables comprehensive genome interpretation and nomination of candidate diagnoses for rare genetic diseases.

BACKGROUND: Clinical interpretation of genetic variants in the context of the patient's phenotype is becoming the largest component of cost and time expenditure for genome-based diagnosis of rare genetic diseases. Artificial intelligence (AI) holds promise to greatly simplify and speed genome interpretation by integrating predictive methods with the growing knowledge of genetic disease. Here we assess the diagnostic performance of Fabric GEM, a new, AI-based, clinical decision support tool for expediting genome interpretation. METHODS: We benchmarked GEM in a retrospective cohort of 119 probands, mostly NICU infants, diagnosed with rare genetic diseases, who received whole-genome or whole-exome sequencing (WGS, WES). We replicated our analyses in a separate cohort of 60 cases collected from five academic medical centers. For comparison, we also analyzed these cases with current state-of-the-art variant prioritization tools. Included in the comparisons were trio, duo, and singleton cases. Variants underpinning diagnoses spanned diverse modes of inheritance and types, including structural variants (SVs). Patient phenotypes were extracted from clinical notes by two means: manually and using an automated clinical natural language processing (CNLP) tool. Finally, 14 previously unsolved cases were reanalyzed. RESULTS: GEM ranked over 90% of the causal genes among the top or second candidate and prioritized for review a median of 3 candidate genes per case, using either manually curated or CNLP-derived phenotype descriptions. Ranking of trios and duos was unchanged when analyzed as singletons. In 17 of 20 cases with diagnostic SVs, GEM identified the causal SVs as the top candidate and in 19/20 within the top five, irrespective of whether SV calls were provided or inferred ab initio by GEM using its own internal SV detection algorithm. GEM showed similar performance in absence of parental genotypes. Analysis of 14 previously unsolved cases resulted in a novel finding for one case, candidates ultimately not advanced upon manual review for 3 cases, and no new findings for 10 cases. CONCLUSIONS: GEM enabled diagnostic interpretation inclusive of all variant types through automated nomination of a very short list of candidate genes and disorders for final review and reporting. In combination with deep phenotyping by CNLP, GEM enables substantial automation of genetic disease diagnosis, potentially decreasing cost and expediting case review.

A data-driven architecture using natural language processing to improve phenotyping efficiency and accelerate genetic diagnoses of rare disorders.

Effective genetic diagnosis requires the correlation of genetic variant data with detailed phenotypic information. However, manual encoding of clinical data into machine-readable forms is laborious and subject to observer bias. Natural language processing (NLP) of electronic health records has great potential to enhance reproducibility at scale but suffers from idiosyncrasies in physician notes and other medical records. We developed methods to optimize NLP outputs for automated diagnosis. We filtered NLP-extracted Human Phenotype Ontology (HPO) terms to more closely resemble manually extracted terms and identified filter parameters across a three-dimensional space for optimal gene prioritization. We then developed a tiered pipeline that reduces manual effort by prioritizing smaller subsets of genes to consider for genetic diagnosis. Our filtering pipeline enabled NLP-based extraction of HPO terms to serve as a sufficient replacement for manual extraction in 92% of prospectively evaluated cases. In 75% of cases, the correct causal gene was ranked higher with our applied filters than without any filters. We describe a framework that can maximize the utility of NLP-based phenotype extraction for gene prioritization and diagnosis. The framework is implemented within a cloud-based modular architecture that can be deployed across health and research institutions.