Search

Using latent variables in gene expression data can help correct unobserved confounders and increase statistical power for expression quantitative trait Loci (eQTL) detection. The probabilistic estimation of expression residuals (PEER) and principal component analysis (PCA) are widely used methods that can remove unwanted variation and improve eQTL discovery power in bulk RNA-seq analysis. However, their performance has not been evaluated extensively in single-cell eQTL analysis, especially for different cell types. Potential challenges arise due to the structure of single-cell RNA-seq data, including sparsity, skewness, and mean-variance relationship. Here, we show by a series of analyses that PEER and PCA require additional quality control and data transformation steps on the pseudo-bulk matrix to obtain valid latent variables; otherwise, it can result in highly correlated factors (Pearson's correlation r = 0.63 ~ 0.99). Incorporating valid PFs/PCs in the eQTL association model would identify 1.7 ~ 13.3% more eGenes. Sensitivity analysis showed that the pattern of change between the number of eGenes detected and fitted PFs/PCs varied significantly in different cell types. In addition, using highly variable genes to generate latent variables could achieve similar eGenes discovery power as using all genes but save considerable computational resources (~ 6.2-fold faster).

Background Trigger finger and carpal tunnel syndrome are the two most common non-traumatic connective tissue disorders of the hand. Both of these conditions frequently co-occur, often in patients with rheumatoid arthritis. However, this phenotypic association is poorly understood. Hypothesising that the co-occurrence of trigger finger and carpal tunnel syndrome might be explained by shared germline predisposition, we aimed to identify a specific genetic locus associated with both diseases. Methods In this genome-wide association study (GWAS), we identified 2908 patients with trigger finger and 436 579 controls from the UK Biobank prospective cohort. We conducted a case-control GWAS for trigger finger, followed by co-localisation analyses with carpal tunnel syndrome summary statistics. To identify putative causal variants and establish their biological relevance, we did fine-mapping analyses and expression quantitative trait loci (eQTL) analyses, using fibroblasts from healthy donors (n=79) and tenosynovium samples from patients with carpal tunnel syndrome (n=77). We conducted a Cox regression for time to trigger finger and carpal tunnel syndrome diagnosis against plasma IGF-1 concentrations in the UK Biobank cohort. Findings Phenome-wide analyses confirmed a marked association between carpal tunnel syndrome and trigger finger in the participants from UK Biobank (odds ratio [OR] 11·97, 95% CI 11·1–13·0; p−300). GWAS for trigger finger identified five independent loci, including one locus, DIRC3, that was co-localised with carpal tunnel syndrome and could be fine-mapped to rs62175241 (0·76, 0·68–0·84; p=5·03 × 10−13). eQTL analyses found a fibroblast-specific association between the protective T allele of rs62175241 and increased DIRC3 and IGFBP5 expression. Increased plasma IGF-1 concentrations were associated with both carpal tunnel syndrome and trigger finger in participants from UK Biobank (hazard ratio >1·04, p Interpretation In this GWAS, the DIRC3 locus on chromosome 2 was significantly associated with both carpal tunnel syndrome and trigger finger, possibly explaining their co-occurrence. The disease-protective allele of rs62175241 was associated with increased expression of long non-coding RNA DIRC3 and its transcriptional target, IGBP5, an antagonist of IGF-1 signalling. These findings suggest a model in which IGF-1 is a driver of both carpal tunnel syndrome and trigger finger, and in which the DIRC3-IGFBP5 axis directly antagonises fibroblastic IGF-1 signalling.

Using latent variables in gene expression data can help correct spurious correlations due to unobserved confounders and increase statistical power for expression Quantitative Trait Loci (eQTL) detection. Probabilistic Estimation of Expression Residuals (PEER) is a widely used statistical method that has been developed to remove unwanted variation and improve eQTL discovery power in bulk RNA-seq analysis. However, its performance has not been largely evaluated in single-cell eQTL data analysis, where it is becoming a commonly used technique. Potential challenges arise due to the structure of single-cell data, including sparsity, skewness, and mean-variance relationship. Here, we show by a series of analyses that this method requires additional quality control and data transformation steps on the pseudo-bulk matrix to obtain valid PEER factors. By using a population-scale single-cell cohort (OneK1K, N = 982), we found that generating PEER factors without further QC or transformation on the pseudo-bulk matrix could result in inferred factors that are highly correlated (Pearson’s correlation r = 0.626∼0.997). Similar spurious correlations were also found in PEER factors inferred from an independent dataset (induced pluripotent stem cells, N = 31). Optimization of the strategy for generating PEER factors and incorporating the improved PEER factors in the eQTL association model can identify 9.0∼23.1% more eQTLs or 1.7%∼13.3% more eGenes. Sensitivity analysis showed that the pattern of change between the number of eGenes detected and PEER factors fitted varied significantly for different cell types. In addition, using highly variable genes (e.g., top 2000) to generate PEER factors could achieve similar eGenes discovery power as using all genes but save considerable computational resources (∼6.2-fold faster). We provide diagnostic guidelines to improve the robustness and avoid potential pitfalls when generating PEER factors for single-cell eQTL association analyses.

Heterogeneity in the clinical presentation of major depressive disorder and response to antidepressants limits clinicians’ ability to accurately predict a specific patient’s eventual response to therapy. Validated depressive symptom profiles may be an important tool for identifying poor outcomes early in the course of treatment. To derive these symptom profiles, we first examined data from 947 depressed subjects treated with selective serotonin reuptake inhibitors (SSRIs) to delineate the heterogeneity of antidepressant response using probabilistic graphical models (PGMs). We then used unsupervised machine learning to identify specific depressive symptoms and thresholds of improvement that were predictive of antidepressant response by 4 weeks for a patient to achieve remission, response, or nonresponse by 8 weeks. Four depressive symptoms (depressed mood, guilt feelings and delusion, work and activities and psychic anxiety) and specific thresholds of change in each at 4 weeks predicted eventual outcome at 8 weeks to SSRI therapy with an average accuracy of 77% (p = 5.5E-08). The same four symptoms and prognostic thresholds derived from patients treated with SSRIs correctly predicted outcomes in 72% (p = 1.25E-05) of 1996 patients treated with other antidepressants in both inpatient and outpatient settings in independent publicly-available datasets. These predictive accuracies were higher than the accuracy of 53% for predicting SSRI response achieved using approaches that (i) incorporated only baseline clinical and sociodemographic factors, or (ii) used 4-week nonresponse status to predict likely outcomes at 8 weeks. The present findings suggest that PGMs providing interpretable predictions have the potential to enhance clinical treatment of depression and reduce the time burden associated with trials of ineffective antidepressants. Prospective trials examining this approach are forthcoming.

Selective serotonin reuptake inhibitors (SSRIs) are standard of care for major depressive disorder (MDD) pharmacotherapy, but only approximately half of these patients remit on SSRI therapy. Our previous genome-wide association study identified a single-nucleotide polymorphism (SNP) signal across the glutamate-rich 3 (ERICH3) gene that was nearly genome-wide significantly associated with plasma serotonin (5-HT) concentrations, which were themselves associated with SSRI response for MDD patients enrolled in the Mayo Clinic PGRN-AMPS SSRI trial. In this study, we performed a meta-analysis which demonstrated that those SNPs were significantly associated with SSRI treatment outcomes in four independent MDD trials. However, the function of ERICH3 and molecular mechanism(s) by which it might be associated with plasma 5-HT concentrations and SSRI clinical response remained unclear. Therefore, we characterized the human ERICH3 gene functionally and identified ERICH3 mRNA transcripts and protein isoforms that are highly expressed in central nervous system cells. Coimmunoprecipitation identified a series of ERICH3 interacting proteins including clathrin heavy chain which are known to play a role in vesicular function. Immunofluorescence showed ERICH3 colocalization with 5-HT in vesicle-like structures, and ERICH3 knock-out dramatically decreased 5-HT staining in SK-N-SH cells as well as 5-HT concentrations in the culture media and cell lysates without changing the expression of 5-HT synthesizing or metabolizing enzymes. Finally, immunofluorescence also showed ERICH3 colocalization with dopamine in human iPSC-derived neurons. These results suggest that ERICH3 may play a significant role in vesicular function in serotonergic and other neuronal cell types, which might help explain its association with antidepressant treatment response.

Recent innovations in droplet-based single-cell RNA-sequencing (scRNA-seq) have provided the technology necessary to investigate biological questions at cellular resolution. With the ability to assay thousands of cells in a single capture, pooling cells from multiple individuals has become a common strategy. Droplets can subsequently be assigned to a specific individual by leveraging their inherent genetic differences, and numerous computational methods have been developed to address this problem. However, another challenge implicit with droplet-based scRNA-seq is the occurrence of doublets - droplets containing two or more cells. The inaccurate assignment of cells to individuals or failure to remove doublets contribute unwanted noise to the data and result in erroneous scientific conclusions. Therefore, it is essential to assign cells to individuals and remove doublets accurately. We present a new framework to improve individual singlet classification and doublet removal through a multi-method intersectional approach.We developed a framework to evaluate the enhancement in donor assignment and doublet removal through the consensus intersection of multiple demultiplexing and doublet detecting methods. The accuracy was assessed using scRNA-seq data of ∼1.4 million peripheral blood mononucleated cells from 1,034 unrelated individuals and ∼90,000 fibroblast cells from 81 unrelated individuals. We show that our approach significantly improves droplet assignment by separating singlets from doublets and classifying the correct individual compared to any single method. We show that the best combination of techniques varies under different biological and experimental conditions, and we present a framework to optimise cell assignment for a given experiment. We offer Demuxafy (https://demultiplexing-doublet-detecting-docs.readthedocs.io/en/latest/index.html) - a framework built-in Singularity to provide clear, consistent documentation of each method and additional tools to simplify and improve demultiplexing and doublet removal. Our results indicate that leveraging multiple demultiplexing and doublet detecting methods improves accuracy and, consequently, downstream analyses in multiplexed scRNA-seq experiments.

We previously reported that testis-specific Y-encoded-like protein (TSPYLs) are transcription regulators for CYP3A4, CYP2C9, and CYP2C19. Here, we observed dual roles for TSPYLs in mediating serotonin transport and the metabolism of selective serotonin reuptake inhibitors (SSRIs) in patients with major depressive disorder (MDD). The widely prescribed SSRIs, citalopram, and escitalopram are metabolized mainly by CYP2C19. The TSPYL1 rs3828743 single nucleotide polymorphism (SNP), which decreases its suppression of CYP2C19 expression, was associated with rapid escitalopram metabolism and worse treatment response in the Mayo PGRN-AMPS clinical trial. We also found that TSPYLs can regulate expression of the serotonin transporter protein, SLC6A4, and, in turn, serotonin transport into cells. The SNPs in tight linkage disequilibrium with the TSPYL1 rs10223646 SNP were significantly correlated with baseline severity of depression in patients with MDD in the Sequenced Treatment Alternatives to Relieve Depression and International SSRI Pharmacogenomics Consortium clinical trials. Our findings suggest that genetic variation in TSPYL genes may be novel indicators for baseline severity of depression and SSRI poor response.

Trigger finger (TF) and carpal tunnel syndrome (CTS) are two common non-traumatic hand disorders that frequently co-occur. By identifying TF and CTS cases in UK Biobank (UKB), we confirmed a highly significant phenotypic association between the diseases. To investigate the genetic basis for this association we performed a genome-wide association study (GWAS) including 2,908 TF cases and 436,579 European controls in UKB, identifying five independent loci. Colocalization with CTS summary statistics identified a co-localized locus at DIRC3 (lncRNA), which was replicated in FinnGen and fine-mapped to rs62175241. Single-cell and bulk eQTL analysis in fibroblasts from healthy donors (n=79) and tenosynovium samples from CTS patients (n=77) showed that the disease-protective rs62175241 allele was associated with increased DIRC3 and IGFBP5 expression. IGFBP5 is a secreted antagonist of IGF-1 signaling, and elevated IGF-1 levels were associated with CTS and TF in UKB, thereby implicating IGF-1 as a driver of both diseases.

The mechanisms by which DNA alleles contribute to disease risk, drug response, and other human phenotypes are highly context-specific, varying across cell types and under different conditions. Human induced pluripotent stem cells (hiPSCs) are uniquely suited to study these context-dependent effects, but to do so requires cell lines from hundreds or potentially thousands of individuals. Village cultures, where multiple hiPSC lines are cultured and differentiated together in a single dish, provide an elegant solution for scaling hiPSC experiments to the necessary sample sizes required for population-scale studies. Here, we show the utility of village models, demonstrating how cells can be assigned back to a donor line using single cell sequencing, and addressing whether line-specific signaling alters the transcriptional profiles of companion lines in a village culture. We generated single cell RNA sequence data from hiPSC lines cultured independently (uni-culture) and in villages at three independent sites. We show that the transcriptional profiles of hiPSC lines are highly consistent between uni- and village cultures for both fresh (0.46 < R < 0.88) and cryopreserved samples (0.46 < R < 0.62). Using a mixed linear model framework, we estimate that the proportion of transcriptional variation across cells is predominantly due to donor effects, with minimal evidence of variation due to culturing in a village system. We demonstrate that the genetic, epigenetic or hiPSC line-specific effects on gene expression are consistent whether the lines are uni- or village-cultured (0.82 < R < 0.94). Finally, we identify the consistency in the landscape of cell states between uni- and village-culture systems. Collectively, we demonstrate that village methods can be effectively used to detect hiPSC line-specific effects including sensitive dynamics of cell states.

Induced pluripotent stem cells generated from patients with geographic atrophy as well as healthy individuals were differentiated to retinal pigment epithelium (RPE) cells. By integrating transcriptional profiles of 127,659 RPE cells generated from 43 individuals with geographic atrophy and 36 controls with genotype data, we identified 439 expression Quantitative Trait (eQTL) loci in cis that were associated with disease status and specific to subpopulations of RPE cells. We identified loci linked to two genes with known associations with geographic atrophy - PILRB and PRPH2, in addition to 43 genes with significant genotype x disease interactions that are candidates for novel genetic associations for geographic atrophy. On a transcriptome-only level, we identified molecular pathways significantly upregulated in geographic atrophy-RPE including in extracellular cellular matrix reorganisation, neurodegeneration, and mitochondrial functions. We subsequently implemented a large-scale proteomics analysis, confirming modification in proteins associated with these pathways. We also identified six significant protein (p) QTL that regulate protein expression in the RPE cells and in geographic atrophy - two of which share variants with cis-eQTL. Transcriptome-wide association analysis identified genes at loci previously associated with age-related macular degeneration. Further analysis conditional on disease status, implicated statistically significant RPE-specific eQTL. This study uncovers important differences in RPE homeostasis associated with geographic atrophy.

Greater than 90% of significant genome-wide association study (GWAS) single-nucleotide polymorphisms (SNPs) are in noncoding regions of the genome, but only 25.6% are known expression quantitative trait loci (eQTLs). Therefore, the function of many significant GWAS SNPs remains unclear. We have identified a novel type of eQTL for which SNPs distant from ligand-activated transcription factor (TF) binding sites can alter target gene expression in a SNP genotype-by-ligand-dependent fashion that we refer to as pharmacogenomic eQTLs (PGx-eQTLs)-loci that may have important pharmacotherapeutic implications. In the present study, we integrated chromatin immunoprecipitation-seq with RNA-seq and SNP genotype data for a panel of lymphoblastoid cell lines to identify 10 novel cis PGx-eQTLs dependent on the ligand-activated TF aryl hydrocarbon receptor (AHR)-a critical environmental sensor for xenobiotic (drug) and immune response. Those 10 cis PGx-eQTLs were eQTLs only after AHR ligand treatment, even though the SNPs did not create/destroy an AHR response element-the DNA sequence motif recognized and bound by AHR. Additional functional studies in multiple cell lines demonstrated that some cis PGx-eQTLs are functional in multiple cell types, whereas others displayed SNP-by-ligand-dependent effects in just one cell type. Furthermore, four of those cis PGx-eQTLs had previously been associated with clinical phenotypes, indicating that those loci might have the potential to inform clinical decisions. Therefore, SNPs across the genome that are distant from TF binding sites for ligand-activated TFs might function as PGx-eQTLs and, as a result, might have important clinical implications for interindividual variation in drug response. SIGNIFICANCE STATEMENT: More than 90% of single-nucleotide polymorphisms (SNPs) that are associated with clinical phenotypes are located in noncoding regions of the genome. However, the mechanisms of action of many of those SNPs have not been elucidated, and drugs may unmask functional expression quantitative trail loci (eQTLs). In the current study, we used drugs that bind to the ligand-activated transcription factor aryl hydrocarbon receptor (AHR) and identified SNPs that were associated with interindividual variation in gene expression following drug exposure-termed pharmacogenomic (PGx)-eQTLs. Possibly of greater significance, those PGx-eQTL SNPs were outside of AHR binding sites, indicating that they do not interrupt AHR DNA recognition. PGx-eQTLs such as those described in this work may have crucial implications for interindividual variation in drug.

Metabolomics provides valuable tools for the study of drug effects, unraveling the mechanism of action and variation in response due to treatment. In this study we used electrochemistry-based targeted metabolomics to gain insights into the mechanisms of action of escitalopram/citalopram focusing on a set of 31 metabolites from neurotransmitter-related pathways. Overall, 290 unipolar patients with major depressive disorder were profiled at baseline, after 4 and 8 weeks of drug treatment. The 17-item Hamilton Depression Rating Scale (HRSD17) scores gauged depressive symptom severity. More significant metabolic changes were found after 8 weeks than 4 weeks post baseline. Within the tryptophan pathway, we noted significant reductions in serotonin (5HT) and increases in indoles that are known to be influenced by human gut microbial cometabolism. 5HT, 5-hydroxyindoleacetate (5HIAA), and the ratio of 5HIAA/5HT showed significant correlations to temporal changes in HRSD17 scores. In the tyrosine pathway, changes were observed in the end products of the catecholamines, 3-methoxy-4-hydroxyphenylethyleneglycol and vinylmandelic acid. Furthermore, two phenolic acids, 4-hydroxyphenylacetic acid and 4-hydroxybenzoic acid, produced through noncanconical pathways, were increased with drug exposure. In the purine pathway, significant reductions in hypoxanthine and xanthine levels were observed. Examination of metabolite interactions through differential partial correlation networks revealed changes in guanosine–homogentisic acid and methionine–tyrosine interactions associated with HRSD17. Genetic association studies using the ratios of these interacting pairs of metabolites highlighted two genetic loci harboring genes previously linked to depression, neurotransmission, or neurodegeneration. Overall, exposure to escitalopram/citalopram results in shifts in metabolism through noncanonical pathways, which suggest possible roles for the gut microbiome, oxidative stress, and inflammation-related mechanisms.

We set out to determine whether machine learning–based algorithms that included functionally validated pharmacogenomic biomarkers joined with clinical measures could predict selective serotonin reuptake inhibitor (SSRI) remission/response in patients with major depressive disorder (MDD). We studied 1,030 white outpatients with MDD treated with citalopram/escitalopram in the Mayo Clinic Pharmacogenomics Research Network Antidepressant Medication Pharmacogenomic Study (PGRN‐AMPS; n = 398), Sequenced Treatment Alternatives to Relieve Depression (STAR*D; n = 467), and International SSRI Pharmacogenomics Consortium (ISPC; n = 165) trials. A genomewide association study for PGRN‐AMPS plasma metabolites associated with SSRI response (serotonin) and baseline MDD severity (kynurenine) identified single nucleotide polymorphisms (SNPs) in DEFB1,ERICH3,AHR, and TSPAN5 that we tested as predictors. Supervised machine‐learning methods trained using SNPs and total baseline depression scores predicted remission and response at 8 weeks with area under the receiver operating curve (AUC) > 0.7 (P 69% (P ≤ 0.07) in STAR*D and ISPC. These results demonstrate that machine learning can achieve accurate and, importantly, replicable prediction of SSRI therapy response using total baseline depression severity combined with pharmacogenomic biomarkers.

CYP2C9 and CYP2C19 are highly polymorphic pharmacogenes; however, clinically actionable genetic variability in drug metabolism due to these genes has been limited to a few common alleles. The identification and functional characterization of less-common open reading frame sequence variation might help to individualize therapy with drugs that are substrates for the enzymes encoded by these genes. The present study identified seven uncharacterized variants each in CYP2C9 and CYP2C19 using next-generation sequence data for 1013 subjects, and functionally characterized the encoded proteins. Constructs were created and transiently expressed in COS-1 cells for the assay of protein concentration and enzyme activities using fluorometric substrates and liquid chromatography- tandem mass spectrometry with tolbutamide (CYP2C9) and (S)-mephenytoin (CYP2C19) as prototypic substrates. The results were compared with the SIFT, Polyphen, and Provean functional prediction software programs. Cytochrome P450 oxidoreductase (CPR) activities were also determined. Positive correlations were observed between protein content and fluorometric enzyme activity for variants of CYP2C9 (P C and CYP2C19 65A>G activities were much lower than predicted based on protein content. Substrate intrinsic clearance values for CYP2C9 218C>T, 343A>C, and CYP2C19 337G>A, 518C>T, 556C>T, and 557G>A were less than 25% of wild-type allozymes. CPR activity levels were similar for all variants. In summary, sequencing of CYP2C9 and CYP2C19 in 1013 subjects identified low-frequency variants that had not previously been functionally characterized. In silico predictions were not always consistent with functional assay results. These observations emphasize the need for high-throughput methods for pharmacogene variant mutagenesis and functional characterization.

The discovery that somatic cells can be reprogrammed to induced pluripotent stem cells (iPSCs) - cells that can be differentiated into any cell type of the three germ layers - has provided a foundation forin vitrohuman disease modelling1,2, drug development1,2, and population genetics studies3,4. In the majority of instances, the expression levels of genes, plays a critical role in contributing to disease risk, or the ability to identify therapeutic targets. However, while the effect of the genetic background of cell lines has been shown to strongly influence gene expression, the effect has not been evaluated at the level of individual cells. Differences in the effect of genetic variation on the gene expression of different cell-types, would provide significant resolution for in vitro research using preprogramed cells. By bringing together single cell RNA sequencing15–21and population genetics, we now have a framework in which to evaluate the cell-types specific effects of genetic variation on gene expression. Here, we performed single cell RNA-sequencing on 64,018 fibroblasts from 79 donors and we mapped expression quantitative trait loci (eQTL) at the level of individual cell types. We demonstrate that the large majority of eQTL detected in fibroblasts are specific to an individual sub-type of cells. To address if the allelic effects on gene expression are dynamic across cell reprogramming, we generated scRNA-seq data in 19,967 iPSCs from 31 reprogramed donor lines. We again identify highly cell type specific eQTL in iPSCs, and show that that the eQTL in fibroblasts are almost entirely disappear during reprogramming. This work provides an atlas of how genetic variation influences gene expression across cell subtypes, and provided evidence for patterns of genetic architecture that lead to cell-types specific eQTL effects.

BackgroundAcylcarnitines have important functions in mitochondrial energetics and β-oxidation, and have been implicated to play a significant role in metabolic functions of the brain. This retrospective study examined whether plasma acylcarnitine profiles can help biochemically distinguish the three phenotypic subtypes of major depressive disorder (MDD)—(core depression (CD+), anxious depression (ANX+), and neurovegetative symptoms of melancholia (NVSM+))—following treatment with a selective serotonin reuptake inhibitor (SSRI).MethodsDepressed outpatients (n=240) from the Mayo Clinic Pharmacogenomics Research Network were treated with citalopram or escitalopram for eight weeks. Plasma samples collected at baseline and eight weeks post-treatment were profiled for multiple-, short-, medium- and long-chain acylcarnitine levels using AbsoluteIDQ®p180-Kit and LC-MS. Linear mixed effects models were used to examine whether acylcarnitine levels discriminate the clinical phenotypes at baseline or eight weeks post-treatment, and whether temporal changes in acylcarnitine profiles differ between groups.ResultsAt baseline, significantly lower concentrations of short- and long-chain acylcarnitines were found in CD+ and NVSM+ compared to ANX+, and the short-chain acylcarnitines remained lower after eight weeks. At eight weeks, the medium- and long-chain acylcarnitines were significantly lower in NVSM+ compared to ANX+. Regarding changes baseline to week eight, short-chain acylcarnitine levels significantly increased in CD+ and ANX+, and medium- and long-chain acylcarnitines significantly decreased in NVSM+ and CD+.ConclusionsIn depressed patients treated with SSRIs, β-oxidation and mitochondrial energetics as evaluated by levels and changes in acylcarnitines may provide the biochemical basis of the clinical heterogeneity of MDD, especially when combined with clinical characteristics.

Millions of patients suffer from major depressive disorder (MDD), but many do not respond to selective serotonin reuptake inhibitor (SSRI) therapy. We used a pharmacometabolomics-informed pharmacogenomics research strategy to identify genes associated with metabolites that were related to SSRI response. Specifically, 306 MDD patients were treated with citalopram or escitalopram and blood was drawn at baseline, 4 and 8 weeks for blood drug levels, genome-wide single nucleotide polymorphism (SNP) genotyping and metabolomic analyses. SSRI treatment decreased plasma serotonin concentrations (P

This work proposes a “learning-augmented clinical assessment” workflow to sequentially augment physician assessments of patients’ symptoms and their socio-demographic measures with heterogeneous biological measures to accurately predict treatment outcomes using machine learning. Across many psychiatric illnesses, ranging from major depressive disorder to schizophrenia, symptom severity assessments are subjective and do not include biological measures, making predictability in eventual treatment outcomes a challenge. Using data from the Mayo Clinic PGRN-AMPS SSRI trial as a case study, this work demonstrates a significant improvement in the prediction accuracy for antidepressant treatment outcomes in patients with major depressive disorder from 35% to 80% individualized by patient, compared to using only a physician’s assessment as the predictors. This improvement is achieved through an iterative overlay of biological measures, starting with metabolites (blood measures modulated by drug action) associated with symptom severity, and then adding in genes associated with metabolomic concentrations. Hence, therapeutic efficacy for a new patient can be assessed prior to treatment, using prediction models that take as inputs, selected biological measures and physician’s assessments of depression severity. Of broader significance extending beyond psychiatry, the approach presented in this work can potentially be applied to predicting treatment outcomes for other medical conditions, such as migraine headaches or rheumatoid arthritis, for which patients are treated according to subject-reported assessments of symptom severity.

Major depressive disorder (MDD) is a heterogeneous disease. Efforts to identify biomarkers for sub-classifying MDD and antidepressant therapy by genome-wide association studies (GWAS) alone have generally yielded disappointing results. We applied a metabolomics-informed genomic research strategy to study the contribution of genetic variation to MDD pathophysiology by assaying 31 metabolites, including compounds from the tryptophan, tyrosine, and purine pathways, in plasma samples from 290 MDD patients. Associations of metabolite concentrations with depressive symptoms were determined, followed by GWAS for selected metabolites and functional validation studies of the genes identified. Kynurenine (KYN), the baseline plasma metabolite that was most highly associated with depressive symptoms, was negatively correlated with severity of those symptoms. GWAS for baseline plasma KYN concentrations identified SNPs across the beta-defensin 1 (DEFB1) and aryl hydrocarbon receptor (AHR) genes that were cis-expression quantitative trait loci (eQTLs) for DEFB1 and AHR mRNA expression, respectively. Furthermore, the DEFB1 locus was associated with severity of MDD symptoms in a larger cohort of 803 MDD patients. Functional studies demonstrated that DEFB1 could neutralize lipopolysaccharide-stimulated expression of KYN-biosynthesizing enzymes in monocytic cells, resulting in altered KYN concentrations in the culture media. In addition, we demonstrated that AHR was involved in regulating the expression of enzymes in the KYN pathway and altered KYN biosynthesis in cell lines of hepatocyte and astrocyte origin. In conclusion, these studies identified SNPs that were cis-eQTLs for DEFB1 and AHR and, which were associated with variation in plasma KYN concentrations that were related to severity of MDD symptoms.

This paper proposes a data-driven longitudinal model that brings together factor graphs and learning methods to demonstrate a significant improvement in predictability in clinical outcomes of patients with major depressive disorder treated with antidepressants. Using data from the Mayo Clinic PGRN-AMPS trial and the STAR∗D trial for validation, this work makes two significant contributions in the context of predictability in psychiatric therapeutic outcomes. First, we establish symptom dynamics in response to antidepressants by using the forward algorithm on a factor graph. Symptom dynamics are the changes in the symptom severity that are most likely to occur because of the antidepressants taken during the trial, and the associated clinical outcomes at 4 weeks and 8 weeks into the trial. The structure of the factor graph is inferred by using unsupervised learning to stratify patients by the similarity of their overall symptom severity. Second, by using metabolomics data as an accurate biological measure in addition to symptom survey data and other patient history information, the prediction of clinical outcomes such as response and remission significantly improved from 30% to 68% in men, and from 35% to 72% in women. This work demonstrates a significant difference in how men and women respond to antidepressants in terms of their symptom dynamics, and also shows that top predictors of clinical outcomes for men and women are significantly different and known to play a role in behavioral sciences.

The aryl hydrocarbon receptor (AHR) is a nuclear receptor that modulates the response to environmental stimuli. It was recognized historically for its role in toxicology but, in recent decades, it has been increasingly recognized as an important modulator of disease—especially for its role in modulating immune and inflammatory responses. AHR has been implicated in many diseases that are driven by immune/inflammatory processes, including major depressive disorder, multiple sclerosis, rheumatoid arthritis, asthma, and allergic responses, among others. The mechanisms by which AHR has been suggested to impact immune/inflammatory diseases include targeted gene expression and altered immune differentiation. It has been suggested that single nucleotide polymorphisms (SNPs) that are near AHR-regulated genes may contribute to AHR-dependent disease mechanisms/pathways. Further, we have found that SNPs that are outside of nuclear receptor binding sites (i.e., outside of AHR response elements (AHREs)) may contribute to AHR-dependent gene regulation in a SNP- and ligand-dependent manner. This review will discuss the evidence and mechanisms of AHR contributions to immune/inflammatory diseases and will consider the possibility that SNPs that are outside of AHR binding sites might contribute to AHR ligand-dependent inter-individual variation in disease pathophysiology and response to pharmacotherapeutics.

The initial decades of the twenty-first century have witnessed striking technical advances that have made it possible to detect, identify and quantitatively measure large numbers of plasma or tissue metabolites. In parallel, similar advances have taken place in our ability to sequence DNA and RNA. Those advances have moved us beyond studies of single metabolites and single genetic polymorphisms to the study of hundreds or thousands of metabolites and millions of genomic variants in a single cell or subject. It is now possible to merge and integrate large data sets generated by the use of different “-omics” techniques to increase our understanding of the molecular basis for variation in disease risk and/or drug response phenotypes. This “Brief Review” will outline some of the challenges and opportunities associated with studies in which metabolomic data have been merged with genomics in an attempt to gain novel insight into mechanisms associated with variation in drug response phenotypes, with an emphasis on the application of a pharmacometabolomics-informed pharmacogenomic research strategy and with selected examples of the application of that strategy. Studies that used pharmacometabolomics to inform and guide pharmacogenomics were reviewed. Clinical studies that were used as the basis for pharmacometabolomics-informed pharmacogenomic studies, published in five independent manuscripts, are described briefly. Within these five manuscripts, both pharmacokinetic and pharmacodynamic metabolomics approaches were used. Candidate gene and genome-wide approaches that were used in concert with these metabolomic data identified novel metabolite-gene relationships that were associated with drug response phenotypes in these pharmacometabolomics-informed pharmacogenomics studies. This “Brief Review” outlines the emerging discipline of pharmacometabolomics-informed pharmacogenomics in which metabolic profiles are associated with both clinical phenotypes and genetic variants to identify novel genetic variants associated with drug response phenotypes based on metabolic profiles.

Major depressive disorder (MDD) is a severe illness that afflicts about 16.6% of people over their lifetime. MDD is highly correlated with suicidality, and often first presents in adolescence. Unfortunately, many pediatric patients suffering from MDD go undiagnosed, and current evidence-based treatment options in the U.S. are limited to psychotherapy and two selective serotonin reuptake inhibitors approved by the United States Food and Drug Administration. Molecular mechanisms have been shown to play a role in MDD pathogenesis, progression, and response to medication, yet few studies have explored the role of these pathways in pediatric MDD. In this review, we outline the gravity and importance of MDD in pediatric patients, some challenges in diagnosis and treatment, current treatments available for pediatric patients, and research to investigate differences between pediatric and adult MDD. We hope that this review will provide an outline of the current understanding and treatment of MDD in pediatric patients, and provide thoughtful insights for future work that could advance our understanding of MDD in pediatric populations, and also identify new therapeutic strategies.

The development, maturation, and maintenance of tissues and organisms are anchored in distinct programs for protein expression which define the identities and roles of individual cell lines [1, 2]. These programs are maintained in a heritable state by epigenetic mecha‐ nisms that convey cellular memory [3, 4]. In this way, the global synchronization of patterns in gene expression broadly dictates developmental consequences [5, 6]. At the foundation of such gene regulation are coordinated cascades that affect the packaging of DNA into chro‐ matin, thereby establishing the degree of DNA accessibility to transcriptional complexes [7-10]. These pathways include histone methylation, methylation of transcriptional regula‐ tors, DNA methylation, histone replacement, chromatin remodeling, and other alterations to histone tails [11-16]. Abnormalities in these epigenetic events are commonly associated with tumorigenesis and subsequent clinical outcomes [17-23].