Clinical implementation of preemptive pharmacogenomics in psychiatry.

BACKGROUND: Pharmacogenomics (PGx) holds promise to revolutionize modern healthcare. Although there are several prospective clinical studies in oncology and cardiology, demonstrating a beneficial effect of PGx-guided treatment in reducing adverse drug reactions, there are very few such studies in psychiatry, none of which spans across all main psychiatric indications, namely schizophrenia, major depressive disorder and bipolar disorder. In this study we aim to investigate the clinical effectiveness of PGx-guided treatment (occurrence of adverse drug reactions, hospitalisations and re-admissions, polypharmacy) and perform a cost analysis of the intervention. METHODS: We report our findings from a multicenter, large-scale, prospective study of pre-emptive genome-guided treatment named as PREemptive Pharmacogenomic testing for preventing Adverse drug REactions (PREPARE) in a large cohort of psychiatric patients (n = 1076) suffering from schizophrenia, major depressive disorder and bipolar disorder. FINDINGS: We show that patients with an actionable phenotype belonging to the PGx-guided arm (n = 25) present with 34.1% less adverse drug reactions compared to patients belonging to the control arm (n = 36), 41.2% less hospitalisations (n = 110 in the PGx-guided arm versus n = 187 in the control arm) and 40.5% less re-admissions (n = 19 in the PGx-guided arm versus n = 32 in the control arm), less duration of initial hospitalisations (n = 3305 total days of hospitalisation in the PGx-guided arm from 110 patients, versus n = 6517 in the control arm from 187 patients) and duration of hospitalisation upon readmission (n = 579 total days of hospitalisation upon readmission in the PGx-guided arm, derived from 19 patients, versus n = 928 in the control arm, from 32 patients respectively). It was also shown that in the vast majority of the cases, there was less drug dose administrated per drug in the PGx-guided arm compared to the control arm and less polypharmacy (n = 124 patients prescribed with at least 4 psychiatric drugs in the PGx-guided arm versus n = 143 in the control arm) and smaller average number of co-administered psychiatric drugs (2.19 in the PGx-guided arm versus 2.48 in the control arm. Furthermore, less deaths were reported in the PGx-guided arm (n = 1) compared with the control arm (n = 9). Most importantly, we observed a 48.5% reduction of treatment costs in the PGx-guided arm with a reciprocal slight increase of the quality of life of patients suffering from major depressive disorder (0.935 versus 0.925 QALYs in the PGx-guided and control arm, respectively). INTERPRETATION: While only a small proportion (∼25%) of the entire study sample had an actionable genotype, PGx-guided treatment can have a beneficial effect in psychiatric patients with a reciprocal reduction of treatment costs. Although some of these findings did not remain significant when all patients were considered, our data indicate that genome-guided psychiatric treatment may be successfully integrated in mainstream healthcare. FUNDING: European Union Horizon 2020.

Development of an optimized and generic cost-utility model for analyzing genome-guided treatment data.

Economic evaluation is an integral component of informed public health decision-making in personalized medicine. However, performing economic evaluation assessments often requires specialized knowledge, expertise, and significant resources. To this end, developing generic models can significantly assist towards providing the necessary evidence for the cost-effectiveness of genome-guided therapeutic interventions, compared to the traditional drug treatment modalities. Here, we report a generic cost-utility analysis model, developed in R, which encompasses essential economic evaluation steps. Specifically, critical steps towards a comprehensive deterministic and probabilistic sensitivity analysis were incorporated in our model, while also providing an easy-to-use graphical user interface, which allows even non-experts in the field to produce a fully comprehensive cost-utility analysis report. To further demonstrate the model's reproducibility, two sets of data were assessed, one stemming from in-house clinical data and one based on previously published data. By implementing the generic model presented herein, we show that the model produces results in complete concordance with the traditionally performed cost-utility analysis for both datasets. Overall, this work demonstrates the potential of generic models to provide useful economic evidence for personalized medicine interventions.

The ethnogeographic variability of genetic factors underlying G6PD deficiency.

Glucose-6-phosphate dehydrogenase (G6PD) deficiency caused by genetic variants in the G6PD gene, constitutes the most common enzymopathy worldwide, affecting approximately 5% of the global population. While carriers are mostly asymptomatic, they are at substantial risk of acute hemolytic anemia upon certain infections or exposure to various medications. As such, information about G6PD activity status in a given patient can constitute an important parameter to guide clinical decision-making. Here, we leveraged whole genome sequencing data from 142,069 unrelated individuals across seven human populations to provide a global comprehensive map of G6PD variability. By integrating established functional classifications with stringent computational predictions using 13 partly orthogonal algorithms for uncharacterized and novel variants, we reveal the large extent of ethnogeographic variability in G6PD deficiency and highlight its population-specific genetic composition. Overall, estimated disease prevalence in males ranged between 12.2% in Africans, 2.7-3.5% across Asia and 2.1% in Middle Easterners to < 0.3% in Europeans, Finnish and Amish. In Africans, the major deficient alleles were A-202A/376 G (minor allele frequency 11.6%) and A-968 C/376 G (0.5%). In contrast, G6PD deficiency in Middle Easterners was primarily due to the Mediterranean allele (1.3%) and the population-specific Cairo variant (0.4%). In South Asia, the most prevalent deficient alleles were Mediterranean (1.7%), Kerala (1.1%), Gond (0.9%) and Orissa (0.2%), whereas in East Asian populations the Canton (1.1%), Kaiping (0.7%) and Viangchan (0.3%) variants were predominant. Combined, our analyses provide a large dataset of G6PD variability across major ethnogeographic groups and can instruct population-specific genotyping strategies to optimize genetically guided therapeutic interventions.

A novel machine learning-based approach for the computational functional assessment of pharmacogenomic variants.

BACKGROUND: The field of pharmacogenomics focuses on the way a person's genome affects his or her response to a certain dose of a specified medication. The main aim is to utilize this information to guide and personalize the treatment in a way that maximizes the clinical benefits and minimizes the risks for the patients, thus fulfilling the promises of personalized medicine. Technological advances in genome sequencing, combined with the development of improved computational methods for the efficient analysis of the huge amount of generated data, have allowed the fast and inexpensive sequencing of a patient's genome, hence rendering its incorporation into clinical routine practice a realistic possibility. METHODS: This study exploited thoroughly characterized in functional level SNVs within genes involved in drug metabolism and transport, to train a classifier that would categorize novel variants according to their expected effect on protein functionality. This categorization is based on the available in silico prediction and/or conservation scores, which are selected with the use of recursive feature elimination process. Toward this end, information regarding 190 pharmacovariants was leveraged, alongside with 4 machine learning algorithms, namely AdaBoost, XGBoost, multinomial logistic regression, and random forest, of which the performance was assessed through 5-fold cross validation. RESULTS: All models achieved similar performance toward making informed conclusions, with RF model achieving the highest accuracy (85%, 95% CI: 0.79, 0.90), as well as improved overall performance (precision 85%, sensitivity 84%, specificity 94%) and being used for subsequent analyses. When applied on real world WGS data, the selected RF model identified 2 missense variants, expected to lead to decreased function proteins and 1 to increased. As expected, a greater number of variants were highlighted when the approach was used on NGS data derived from targeted resequencing of coding regions. Specifically, 71 variants (out of 156 with sufficient annotation information) were classified as to "Decreased function," 41 variants as "No" function proteins, and 1 variant in "Increased function." CONCLUSION: Overall, the proposed RF-based classification model holds promise to lead to an extremely useful variant prioritization and act as a scoring tool with interesting clinical applications in the fields of pharmacogenomics and personalized medicine.

A Novel Text-Mining Approach for Retrieving Pharmacogenomics Associations From the Literature.

Text mining in biomedical literature is an emerging field which has already been shown to have a variety of implementations in many research areas, including genetics, personalized medicine, and pharmacogenomics. In this study, we describe a novel text-mining approach for the extraction of pharmacogenomics associations. The code that was used toward this end was implemented using R programming language, either through custom scripts, where needed, or through utilizing functions from existing libraries. Articles (abstracts or full texts) that correspond to a specified query were extracted from PubMed, while concept annotations were derived by PubTator Central. Terms that denote a Mutation or a Gene as well as Chemical compound terms corresponding to drug compounds were normalized and the sentences containing the aforementioned terms were filtered and preprocessed to create appropriate training sets. Finally, after training and adequate hyperparameter tuning, four text classifiers were created and evaluated (FastText, Linear kernel SVMs, XGBoost, Lasso, and Elastic-Net Regularized Generalized Linear Models) with regard to their performance in identifying pharmacogenomics associations. Although further improvements are essential toward proper implementation of this text-mining approach in the clinical practice, our study stands as a comprehensive, simplified, and up-to-date approach for the identification and assessment of research articles enriched in clinically relevant pharmacogenomics relationships. Furthermore, this work highlights a series of challenges concerning the effective application of text mining in biomedical literature, whose resolution could substantially contribute to the further development of this field.

Exome-Wide Analysis of the DiscovEHR Cohort Reveals Novel Candidate Pharmacogenomic Variants for Clinical Pharmacogenomics.

Recent advances in next-generation sequencing technology have led to the production of an unprecedented volume of genomic data, thus further advancing our understanding of the role of genetic variation in clinical pharmacogenomics. In the present study, we used whole exome sequencing data from 50,726 participants, as derived from the DiscovEHR cohort, to identify pharmacogenomic variants of potential clinical relevance, according to their occurrence within the PharmGKB database. We further assessed the distribution of the identified rare and common pharmacogenomics variants amongst different GnomAD subpopulations. Overall, our findings show that the use of publicly available sequence data, such as the DiscovEHR dataset and GnomAD, provides an opportunity for a deeper understanding of genetic variation in pharmacogenes with direct implications in clinical pharmacogenomics.

Documentation of clinically relevant genomic biomarker allele frequencies in the next-generation FINDbase worldwide database.

FINDbase (http://www.findbase.org) is a comprehensive data resource recording the prevalence of clinically relevant genomic variants in various populations worldwide, such as pathogenic variants underlying genetic disorders as well as pharmacogenomic biomarkers that can guide drug treatment. Here, we report significant new developments and technological advancements in the database architecture, leading to a completely revamped database structure, querying interface, accompanied with substantial extensions of data content and curation. In particular, the FINDbase upgrade further improves the user experience by introducing responsive features that support a wide variety of mobile and stationary devices, while enhancing computational runtime due to the use of a modern Javascript framework such as ReactJS. Data collection is significantly enriched, with the data records being divided in a Public and Private version, the latter being accessed on the basis of data contribution, according to the microattribution approach, while the front end was redesigned to support the new functionalities and querying tools. The abovementioned updates further enhance the impact of FINDbase, improve the overall user experience, facilitate further data sharing by microattribution, and strengthen the role of FINDbase as a key resource for personalized medicine applications and personalized public health.

Rethinking Drug Repositioning and Development with Artificial Intelligence, Machine Learning, and Omics.

Pharmaceutical industry and the art and science of drug development are sorely in need of novel transformative technologies in the current age of digital health and artificial intelligence (AI). Often described as game-changing technologies, AI and machine learning algorithms have slowly but surely begun to revolutionize pharmaceutical industry and drug development over the past 5 years. In this expert review, we describe the most frequently used machine learning algorithms in drug development pipelines and the -omics databases well poised to support machine learning and drug discovery. Subsequently, we analyze the emerging new computational approaches to drug discovery and the in silico pipelines for drug repositioning and the synergies among -omics system sciences, AI and machine learning. As with system sciences, AI and machine learning embody a system scale and Big Data driven vision for drug discovery and development. We conclude with a future outlook on the ways in which machine learning approaches can be implemented to buttress and expedite drug discovery and precision medicine. As AI and machine learning are rapidly entering pharmaceutical industry and the art and science of drug development, we need to critically examine the attendant prospects and challenges to benefit patients and public health.