A primer on the use of machine learning to distil knowledge from data in biological psychiatry.

Applications of machine learning in the biomedical sciences are growing rapidly. This growth has been spurred by diverse cross-institutional and interdisciplinary collaborations, public availability of large datasets, an increase in the accessibility of analytic routines, and the availability of powerful computing resources. With this increased access and exposure to machine learning comes a responsibility for education and a deeper understanding of its bases and bounds, borne equally by data scientists seeking to ply their analytic wares in medical research and by biomedical scientists seeking to harness such methods to glean knowledge from data. This article provides an accessible and critical review of machine learning for a biomedically informed audience, as well as its applications in psychiatry. The review covers definitions and expositions of commonly used machine learning methods, and historical trends of their use in psychiatry. We also provide a set of standards, namely Guidelines for REporting Machine Learning Investigations in Neuropsychiatry (GREMLIN), for designing and reporting studies that use machine learning as a primary data-analysis approach. Lastly, we propose the establishment of the Machine Learning in Psychiatry (MLPsych) Consortium, enumerate its objectives, and identify areas of opportunity for future applications of machine learning in biological psychiatry. This review serves as a cautiously optimistic primer on machine learning for those on the precipice as they prepare to dive into the field, either as methodological practitioners or well-informed consumers.

Family-centred care interventions for children with chronic conditions: A scoping review.

INTRODUCTION: Children with chronic conditions have greater health care needs than the general paediatric population but may not receive care that centres their needs and preferences as identified by their families. Clinicians and researchers are interested in developing interventions to improve family-centred care need information about the characteristics of existing interventions, their development and the domains of family-centred care that they address. We conducted a scoping review that aimed to identify and characterize recent family-centred interventions designed to improve experiences with care for children with chronic conditions. METHODS: We searched Medline, Embase, PsycInfo and Cochrane databases, and grey literature sources for relevant articles or documents published between 1 January 2019 and 11 August 2020 (databases) or 7-20 October 2020 (grey literature). Primary studies with ≥10 participants, clinical practice guidelines and theoretical articles describing family-centred interventions that aimed to improve experiences with care for children with chronic conditions were eligible. Following citation and full-text screening by two reviewers working independently, we charted data covering study characteristics and interventions from eligible reports and synthesized interventions by domains of family-centred care. RESULTS: Our search identified 2882 citations, from which 63 articles describing 61 unique interventions met the eligibility criteria and were included in this review. The most common study designs were quasiexperimental studies (n = 18), randomized controlled trials (n = 11) and qualitative and mixed-methods studies (n = 9 each). The most frequently addressed domains of family-centred care were communication and information provision (n = 45), family involvement in care (n = 37) and access to care (n = 30). CONCLUSION: This review, which identified 61 unique interventions aimed at improving family-centred care for children with chronic conditions across a range of settings, is a concrete resource for researchers, health care providers and administrators interested in improving care for this high-needs population. PATIENT OR PUBLIC CONTRIBUTION: This study was co-developed with three patient partner co-investigators, all of whom are individuals with lived experiences of rare chronic diseases as parents and/or patients and have prior experience in patient engagement in research (I. J., N. P., M. S.). These patient partner co-investigators contributed to this study at all stages, from conceptualization to dissemination.

Assessing the quality and value of metabolic chart data for capturing core outcomes for pediatric medium-chain acyl-CoA dehydrogenase (MCAD) deficiency.

BACKGROUND: Generating rigorous evidence to inform care for rare diseases requires reliable, sustainable, and longitudinal measurement of priority outcomes. Having developed a core outcome set for pediatric medium-chain acyl-CoA dehydrogenase (MCAD) deficiency, we aimed to assess the feasibility of prospective measurement of these core outcomes during routine metabolic clinic visits. METHODS: We used existing cohort data abstracted from charts of 124 children diagnosed with MCAD deficiency who participated in a Canadian study which collected data from birth to a maximum of 11 years of age to investigate the frequency of clinic visits and quality of metabolic chart data for selected outcomes. We recorded all opportunities to collect outcomes from the medical chart as a function of visit rate to the metabolic clinic, by treatment centre and by child age. We applied a data quality framework to evaluate data based on completeness, conformance, and plausibility for four core MCAD outcomes: emergency department use, fasting time, metabolic decompensation, and death. RESULTS: The frequency of metabolic clinic visits decreased with increasing age, from a rate of 2.8 visits per child per year (95% confidence interval, 2.3-3.3) among infants 2 to 6 months, to 1.0 visit per child per year (95% confidence interval, 0.9-1.2) among those ≥ 5 years of age. Rates of emergency department visits followed anticipated trends by child age. Supplemental findings suggested that some emergency visits occur outside of the metabolic care treatment centre but are not captured. Recommended fasting times were updated relatively infrequently in patients' metabolic charts. Episodes of metabolic decompensation were identifiable but required an operational definition based on acute manifestations most commonly recorded in the metabolic chart. Deaths occurred rarely in these patients and quality of mortality data was not evaluated. CONCLUSIONS: Opportunities to record core outcomes at the metabolic clinic occur at least annually for children with MCAD deficiency. Methods to comprehensively capture emergency care received at outside institutions are needed. To reduce substantial heterogeneous recording of core outcome across treatment centres, improved documentation standards are required for recording of recommended fasting times and a consensus definition for metabolic decompensations needs to be developed and implemented.

An in-frame deletion affecting the critical acid loop of PPP2R5D is associated with a neonatal lethal form of PPP2R5D-related neurodevelopmental disorder.

Heterozygous pathogenic variants in PPP2R5D gene are associated with PPP2R5D-related neurodevelopmental disorder, a rare autosomal dominant condition, characterized by neurodevelopmental impairment in childhood, macrocephaly/megalencephaly, hypotonia, epilepsy, and dysmorphic features. Up-to-date, only approximately 100 cases have been published in the literature and the full phenotypic and genotypic spectrum have not yet been fully described. PPP2R5D gene encodes the B56δ subunit of the PP2A enzyme complex. We describe a neonatal form of PPP2R5D-related disorder with early infantile death, caused by a novel in-frame deletion causing loss of 8 amino acids and insertion of serine at position 201 (p.Phe194_Pro201delinsSer) of the B56δ subunit. This deletion is predicted to disrupt a critical acidic loop of amino acids important for binding other subunits of the PP2A enzyme complex, and harbors many of the residues previously reported to cause a mild-moderate form of this condition. This report describes a neonatal lethal presentation of the PPP2R5D-related neurodevelopmental disorder and provides additional evidence that disruption of the acidic loop is an important pathomechanism underlying PPP2R5D-related disorder.

Ethical and practical considerations related to data sharing when collecting patient-reported outcomes in care-based child health research.

PURPOSE: The collection and use of patient reported outcomes (PROs) in care-based child health research raises challenging ethical and logistical questions. This paper offers an analysis of two questions related to PROs in child health research: (1) Is it ethically obligatory, desirable or preferable to share PRO data collected for research with children, families, and health care providers? And if so, (2) What are the characteristics of a model best suited to guide the collection, monitoring, and sharing of these data? METHODS: A multidisciplinary team of researchers, providers, patient and family partners, and ethicists examined the literature and identified a need for focus on PRO sharing in pediatric care-based research. We constructed and analyzed three models for managing pediatric PRO data in care-based research, drawing on ethical principles, logistics, and opportunities to engage with children and families. RESULTS: We argue that it is preferable to share pediatric PRO data with providers, but to manage expectations and balance the risks and benefits of research, this requires a justifiable data sharing model. We argue that a successful PRO data sharing model will allow children and families to have access to and control over their own PRO data and be engaged in decision-making around how PROs collected for research may be integrated into care, but require support from providers. CONCLUSION: We propose a PRO data sharing model that can be used across diverse research settings and contributes to improved transparency, communication, and patient-centered care and research.

Machine learning and bioinformatic analysis of brain and blood mRNA profiles in major depressive disorder: A case-control study.

This study analyzed gene expression messenger RNA data, from cases with major depressive disorder (MDD) and controls, using supervised machine learning (ML). We built on the methodology of prior studies to obtain more generalizable/reproducible results. First, we obtained a classifier trained on gene expression data from the dorsolateral prefrontal cortex of post-mortem MDD cases (n = 126) and controls (n = 103). An average area-under-the-receiver-operating-characteristics-curve (AUC) from 10-fold cross-validation of 0.72 was noted, compared to an average AUC of 0.55 for a baseline classifier (p = .0048). The classifier achieved an AUC of 0.76 on a previously unused testing-set. We also performed external validation using DLPFC gene expression values from an independent cohort of matched MDD cases (n = 29) and controls (n = 29), obtained from Affymetrix microarray (vs. Illumina microarray for the original cohort) (AUC: 0.62). We highlighted gene sets differentially expressed in MDD that were enriched for genes identified by the ML algorithm. Next, we assessed the ML classification performance in blood-based microarray gene expression data from MDD cases (n = 1,581) and controls (n = 369). We observed a mean AUC of 0.64 on 10-fold cross-validation, which was significantly above baseline (p = .0020). Similar performance was observed on the testing-set (AUC: 0.61). Finally, we analyzed the classification performance in covariates subgroups. We identified an interesting interaction between smoking and recall performance in MDD case prediction (58% accurate predictions in cases who are smokers vs. 43% accurate predictions in cases who are non-smokers). Overall, our results suggest that ML in combination with gene expression data and covariates could further our understanding of the pathophysiology in MDD.

Machine Learning Analysis of Blood microRNA Data in Major Depression: A Case-Control Study for Biomarker Discovery.

BACKGROUND: There is a lack of reliable biomarkers for major depressive disorder (MDD) in clinical practice. However, several studies have shown an association between alterations in microRNA levels and MDD, albeit none of them has taken advantage of machine learning (ML). METHOD: Supervised and unsupervised ML were applied to blood microRNA expression profiles from a MDD case-control dataset (n = 168) to distinguish between (1) case vs control status, (2) MDD severity levels defined based on the Montgomery-Asberg Depression Rating Scale, and (3) antidepressant responders vs nonresponders. RESULTS: MDD cases were distinguishable from healthy controls with an area-under-the receiver-operating characteristic curve (AUC) of 0.97 on testing data. High- vs low-severity cases were distinguishable with an AUC of 0.63. Unsupervised clustering of patients, before supervised ML analysis of each cluster for MDD severity, improved the performance of the classifiers (AUC of 0.70 for cluster 1 and 0.76 for cluster 2). Antidepressant responders could not be successfully separated from nonresponders, even after patient stratification by unsupervised clustering. However, permutation testing of the top microRNA, identified by the ML model trained to distinguish responders vs nonresponders in each of the 2 clusters, showed an association with antidepressant response. Each of these microRNA markers was only significant when comparing responders vs nonresponders of the corresponding cluster, but not using the heterogeneous unclustered patient set. CONCLUSIONS: Supervised and unsupervised ML analysis of microRNA may lead to robust biomarkers for monitoring clinical evolution and for more timely assessment of treatment in MDD patients.

Machine learning analysis of exome trios to contrast the genomic architecture of autism and schizophrenia.

BACKGROUND: Machine learning (ML) algorithms and methods offer great tools to analyze large complex genomic datasets. Our goal was to compare the genomic architecture of schizophrenia (SCZ) and autism spectrum disorder (ASD) using ML. METHODS: In this paper, we used regularized gradient boosted machines to analyze whole-exome sequencing (WES) data from individuals SCZ and ASD in order to identify important distinguishing genetic features. We further demonstrated a method of gene clustering to highlight which subsets of genes identified by the ML algorithm are mutated concurrently in affected individuals and are central to each disease (i.e., ASD vs. SCZ "hub" genes). RESULTS: In summary, after correcting for population structure, we found that SCZ and ASD cases could be successfully separated based on genetic information, with 86-88% accuracy on the testing dataset. Through bioinformatic analysis, we explored if combinations of genes concurrently mutated in patients with the same condition ("hub" genes) belong to specific pathways. Several themes were found to be associated with ASD, including calcium ion transmembrane transport, immune system/inflammation, synapse organization, and retinoid metabolic process. Moreover, ion transmembrane transport, neurotransmitter transport, and microtubule/cytoskeleton processes were highlighted for SCZ. CONCLUSIONS: Our manuscript introduces a novel comparative approach for studying the genetic architecture of genetically related diseases with complex inheritance and highlights genetic similarities and differences between ASD and SCZ.

Machine learning in schizophrenia genomics, a case-control study using 5,090 exomes.

Our hypothesis is that machine learning (ML) analysis of whole exome sequencing (WES) data can be used to identify individuals at high risk for schizophrenia (SCZ). This study applies ML to WES data from 2,545 individuals with SCZ and 2,545 unaffected individuals, accessed via the database of genotypes and phenotypes (dbGaP). Single nucleotide variants and small insertions and deletions were annotated by ANNOVAR using the reference genome hg19/GRCh37. Rare (predicted functional) variants with a minor allele frequency ≤1% and genotype quality ≥90 including missense, frameshift, stop gain, stop loss, intronic, and exonic splicing variants were selected. A file containing all cases and controls, the names of genes with variants meeting our criteria, and the number of variants per gene for each individual, was used for ML analysis. The supervised machine-learning algorithm used the patterns of variants observed in the different genes to determine which subset of genes can best predict that an individual is affected. Seventy percent of the data was used to train the algorithm and the remaining 30% of data (n = 1,526) was used to evaluate its efficiency. The supervised ML algorithm, gradient boosted trees with regularization (eXtreme Gradient Boosting implementation) was the best performing algorithm yielding promising results (accuracy: 85.7%, specificity: 86.6%, sensitivity: 84.9%, area under the receiver-operator characteristic curve: 0.95). The top 50 features (genes) of the algorithm were analyzed using bioinformatics resources for new insights about the pathophysiology of SCZ. This manuscript presents a novel predictor which could potentially enable studies exploring disease-modifying intervention in the early stages of the disease.

Biomarkers for major depressive and bipolar disorders using metabolomics: A systematic review.

Major depressive disorder (MDD) and bipolar disorder (BD) lack robust biomarkers useful for screening purposes in a clinical setting. A systematic review of the literature was conducted on metabolomic studies of patients with MDD or BD through the use of analytical platforms such as in vivo brain imaging, mass spectrometry, and nuclear magnetic resonance. Our search identified a total of 7,590 articles, of which 266 articles remained for full-text revision. Overall, 249 metabolites were found to be dysregulated with 122 of these metabolites being reported in two or more of the studies included. A list of biomarkers for MDD and BD established from metabolites found to be abnormal, along with the number of studies supporting each metabolite and a comparison of which biological fluids they were reported in, is provided. Metabolic pathways that may be important in the pathophysiology of MDD and BD were identified and predominantly center on glutamatergic metabolism, energy metabolism, and neurotransmission. Using online drug registries, we also illustrate how metabolomics can facilitate the discovery of novel candidate drug targets.

Inborn errors of metabolism associated with psychosis: literature review and case-control study using exome data from 5090 adult individuals.

A literature review was conducted, using the computerized "Online Mendelian Inheritance in Man" (OMIM) and PubMed, to identify inborn errors of metabolism (IEM) in which psychosis may be a predominant feature or the initial presenting symptom. Different combinations of the following keywords were searched using OMIM: "psychosis", "schizophrenia", or "hallucinations" and "metabolic", "inborn error of metabolism", "inborn errors of metabolism", "biochemical genetics", or "metabolic genetics". The OMIM search generated 126 OMIM entries, 40 of which were well known IEM. After removing IEM lacking evidence in PubMed for an association with psychosis, 29 OMIM entries were identified. Several of these IEM are treatable. They involve different small organelles (lysosomes, peroxisomes, mitochondria), iron or copper accumulation, as well as defects in other met-abolic pathways (e.g., defects leading to hyperammonemia or homocystinemia). A clinical checklist summarizing the key features of these conditions and a guide to clinical approach are provided. The genes corresponding to each of these con-ditions were identified. Whole exome data from 2545 adult cases with schizophrenia and 2545 unrelated controls, accessed via the Database of Genotypes and Phenotypes (dbGaP), were analyzed for rare functional variants in these genes. The odds ratio of having a rare functional variant in cases versus controls was calculated for each gene. Eight genes are significantly associated with schizophrenia (p < 0.05, OR >1) using an unselected group of adult patients with schizophrenia. Increased awareness of clinical clues for these IEM will optimize referrals and timely metabolic interventions.

Metabolomics in patients with psychosis: A systematic review.

The purpose of this article is to provide a comprehensive review of metabolomics studies for psychosis, as a means of biomarker discovery. Manuscripts were selected for review if they involved discovery of metabolites using high-throughput analysis in human subjects and were published in the last decade. The metabolites identified were searched in Human Metabolome Data Base (HMDB) for a link to psychosis. Metabolites associated with psychosis based on evidence in HMBD were then searched using PubMed to explore the availability of further evidence. Almost all of the studies which underwent full review involved patients with schizophrenia. Ten biomarkers were identified. Six of them were reported in two or more independent metabolomics studies: N-acetyl aspartate, lactate, tryptophan, kynurenine, glutamate, and creatine. Four additional metabolites were encountered in a single metabolomics study but had significant evidence (two supporting articles or more) for a link to psychosis based on PubMed: linoleic acid, D-serine, glutathione, and 3-hydroxybutyrate. The pathways affected are discussed as they may be relevant to the pathophysiology of psychosis, and specifically of schizophrenia, as well as, constitute new drug targets for treatment of related conditions. Based on the biomarkers identified, early diagnosis of schizophrenia and/or monitoring may be possible.

Prevalence of four Mendelian disorders associated with autism in 2392 affected families.

Autism spectrum disorder (ASD) is a neurobehavioral disorder with a heterogeneous genetic etiology. Based on the literature, several single-gene disorders, including Rett syndrome, Smith-Lemli-Opitz syndrome, PTEN hamartoma tumor syndrome and tuberous sclerosis, are associated with a high prevalence of ASD. We estimated the prevalence of these four conditions in a large cohort of patients using whole-exome sequencing data from 2392 families (1800 quads and 592 trios) with ASD from the National Database for Autism Research. Seven patients carried a pathogenic or likely pathogenic variant in either TSC1, TSC2, PTEN, DHCR7 or MECP2, with 6 out of 7 reportable variants occurring in PTEN (1 in 399).

Child and family experiences with inborn errors of metabolism: a qualitative interview study with representatives of patient groups.

BACKGROUND: Patient-centered health care for children with inborn errors of metabolism (IEM) and their families is important and requires an understanding of patient experiences, needs, and priorities. IEM-specific patient groups have emerged as important voices within these rare disease communities and are uniquely positioned to contribute to this understanding. We conducted qualitative interviews with IEM patient group representatives to increase understanding of patient and family experiences, needs, and priorities and inform patient-centered research and care. METHODS: We developed a sampling frame of patient groups representing IEM disease communities from Canada, the United States, and United Kingdom. With consent, we interviewed participants to explore their views on experiences, needs, and outcomes that are most important to children with IEM and their families. We analyzed the data using a qualitative descriptive approach to identify key themes and sub-themes. RESULTS: We interviewed 18 organizational representatives between February 28 and September 17, 2014, representing 16 IEMs and/or disease categories. Twelve participants voluntarily self-identified as parents and/or were themselves patients. Three key themes emerged from the coded data: managing the uncertainty associated with raising and caring for a child with a rare disease; challenges associated with the affected child's life transitions, and; the collective struggle for improved outcomes and interventions that rare disease communities navigate. CONCLUSION: Health care providers can support children with IEM and their families by acknowledging and reducing uncertainty, supporting families through children's life transitions, and contributing to rare disease communities' progress toward improved interventions, experiences, and outcomes.

Scoping review of patient- and family-oriented outcomes and measures for chronic pediatric disease.

BACKGROUND: Improvements in health care for children with chronic diseases must be informed by research that emphasizes outcomes of importance to patients and families. To support a program of research in the field of rare inborn errors of metabolism (IEM), we conducted a broad scoping review of primary studies that: (i) focused on chronic pediatric diseases similar to IEM in etiology or manifestations and in complexity of management; (ii) reported patient- and/or family-oriented outcomes; and (iii) measured these outcomes using self-administered tools. METHODS: We developed a comprehensive review protocol and implemented an electronic search strategy to identify relevant citations in Medline, EMBASE, DARE and Cochrane. Two reviewers applied pre-specified criteria to titles/abstracts using a liberal accelerated approach. Articles eligible for full-text review were screened by two independent reviewers with discrepancies resolved by consensus. One researcher abstracted data on study characteristics, patient- and family-oriented outcomes, and self-administered measures. Data were validated by a second researcher. RESULTS: 4,118 citations were screened with 304 articles included. Across all included reports, the most-represented diseases were diabetes (35%), cerebral palsy (23%) and epilepsy (18%). We identified 43 unique patient- and family-oriented outcomes from among five emergent domains, with mental health outcomes appearing most frequently. The studies reported the use of 405 independent self-administered measures of these outcomes. CONCLUSIONS: Patient- and family-oriented research investigating chronic pediatric diseases emphasizes mental health and appears to be relatively well-developed in the diabetes literature. Future research can build on this foundation while identifying additional outcomes that are priorities for patients and families.

Association of novel mutation in TRPV4 with familial nonsyndromic craniosynostosis with complete penetrance and variable expressivity.

OBJECTIVE: The aim of this study was to characterize a novel pathogenic variant in the transient receptor potential vanilloid 4 (TRPV4) gene, causing familial nonsyndromic craniosynostosis (CS) with complete penetrance and variable expressivity. METHODS: Whole-exome sequencing was performed on germline DNA of a family with nonsyndromic CS to a mean depth coverage of 300× per sample, with greater than 98% of the targeted region covered at least 25×. In this study, the authors detected a novel variant, c.496C>A in TRPV4, exclusively in the four affected family members. The variant was modeled using the structure of the TRPV4 protein from Xenopus tropicalis. In vitro assays in HEK293 cells overexpressing wild-type TRPV4 or TRPV4 p.Leu166Met were used to assess the effect of the mutation on channel activity and downstream MAPK signaling. RESULTS: The authors identified a novel, highly penetrant heterozygous variant in TRPV4 (NM_021625.4:c.496C>A) causing nonsyndromic CS in a mother and all three of her children. This variant results in an amino acid change (p.Leu166Met) in the intracellular ankyrin repeat domain distant from the Ca2+-dependent membrane channel domain. In contrast to other TRPV4 mutations in channelopathies, this variant does not interfere with channel activity as identified by in silico modeling and in vitro overexpression assays in HEK293 cells. CONCLUSIONS: Based on these findings, the authors hypothesized that this novel variant causes CS by modulating the binding of allosteric regulatory factors to TRPV4 rather than directly modifying its channel activity. Overall, this study expands the genetic and functional spectrum of TRPV4 channelopathies and is particularly relevant for the genetic counseling of CS patients.

Paper II: thematic framework analysis of registry-based randomized controlled trials provided insights for designing trial ready registries.

OBJECTIVES: Registry-based randomized controlled trials (RRCTs) are increasingly used, promising to address challenges associated with traditional randomized controlled trials. We identified strengths and limitations reported in planned and completed RRCTs to inform future RRCTs. STUDY DESIGN AND SETTING: We conducted an environmental scan of literature discussing conceptual or methodological strengths and limitations of using registries for trial design and conduct (n = 12), followed by an analysis of RRCT protocols (n = 13) and reports (n = 77) identified from a scoping review. Using framework analysis, we developed and refined a conceptual framework of RRCT-specific strengths and limitations. We mapped and interpreted strengths and limitations discussed by authors of RRCT articles using framework codes and quantified the frequencies at which these were mentioned. RESULTS: Our conceptual framework identified six main RRCT strengths and four main RRCT limitations. Considering implications for RRCT conduct and design, we formulated ten recommendations for registry designers, administrators, and trialists planning future RRCTs. CONCLUSION: Consideration and application of empirically underpinned recommendations for future registry design and trial conduct may help trialists utilize registries and RRCTs to their full potential.

PSMC3 proteasome subunit variants are associated with neurodevelopmental delay and type I interferon production.

A critical step in preserving protein homeostasis is the recognition, binding, unfolding, and translocation of protein substrates by six AAA-ATPase proteasome subunits (ATPase-associated with various cellular activities) termed PSMC1-6, which are required for degradation of proteins by 26S proteasomes. Here, we identified 15 de novo missense variants in the PSMC3 gene encoding the AAA-ATPase proteasome subunit PSMC3/Rpt5 in 23 unrelated heterozygous patients with an autosomal dominant form of neurodevelopmental delay and intellectual disability. Expression of PSMC3 variants in mouse neuronal cultures led to altered dendrite development, and deletion of the PSMC3 fly ortholog Rpt5 impaired reversal learning capabilities in fruit flies. Structural modeling as well as proteomic and transcriptomic analyses of T cells derived from patients with PSMC3 variants implicated the PSMC3 variants in proteasome dysfunction through disruption of substrate translocation, induction of proteotoxic stress, and alterations in proteins controlling developmental and innate immune programs. The proteostatic perturbations in T cells from patients with PSMC3 variants correlated with a dysregulation in type I interferon (IFN) signaling in these T cells, which could be blocked by inhibition of the intracellular stress sensor protein kinase R (PKR). These results suggest that proteotoxic stress activated PKR in patient-derived T cells, resulting in a type I IFN response. The potential relationship among proteosome dysfunction, type I IFN production, and neurodevelopment suggests new directions in our understanding of pathogenesis in some neurodevelopmental disorders.

Transcriptomics and machine learning to advance schizophrenia genetics: A case-control study using post-mortem brain data.

BACKGROUND AND OBJECTIVE: Alterations of the expression of a variety of genes have been reported in patients with schizophrenia (SCZ). Moreover, machine learning (ML) analysis of gene expression microarray data has shown promising preliminary results in the study of SCZ. Our objective was to evaluate the performance of ML in classifying SCZ cases and controls based on gene expression microarray data from the dorsolateral prefrontal cortex. METHODS: We apply a state-of-the-art ML algorithm (XGBoost) to train and evaluate a classification model using 201 SCZ cases and 278 controls. We utilized 10-fold cross-validation for model selection, and a held-out testing set to evaluate the model. The performance metric utilizes to evaluate classification performance was the area under the receiver-operator characteristics curve (AUC). RESULTS: We report an average AUC on 10-fold cross-validation of 0.76 and an AUC of 0.76 on testing data, not used during training. Analysis of the rolling balanced classification accuracy from high to low prediction confidence levels showed that the most certain subset of predictions ranged between 80-90%. The ML model utilized 182 gene expression probes. Further improvement to classification performance was observed when applying an automated ML strategy on the 182 features, which achieved an AUC of 0.79 on the same testing data. We found literature evidence linking all of the top ten ML ranked genes to SCZ. Furthermore, we leveraged information from the full set of microarray gene expressions available via univariate differential gene expression analysis. We then prioritized differentially expressed gene sets using the piano gene set analysis package. We augmented the ranking of the prioritized gene sets with genes from the complex multivariate ML model using hypergeometric tests to identify more robust gene sets. We identified two significant Gene Ontology molecular function gene sets: "oxidoreductase activity, acting on the CH-NH2 group of donors" and "integrin binding." Lastly, we present candidate treatments for SCZ based on findings from our study CONCLUSIONS: Overall, we observed above-chance performance from ML classification of SCZ cases and controls based on brain gene expression microarray data, and found that ML analysis of gene expressions could further our understanding of the pathophysiology of SCZ and help identify novel treatments.