Methods to analyze big data in pharmacogenomics research

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.

Considerations When Applying Pharmacogenomics to Your Practice

Many practitioners who have not had pharmacogenomic education are required to apply pharmacogenomics to their practices. Although many aspects of pharmacogenomics are similar to traditional concepts of drug-drug interactions, there are some differences. We searched PubMed with the search terms pharmacogenomics and pharmacogenetics (January 1, 2005, through December 31, 2019) and selected articles that supported the application of pharmacogenomics to practice. For inclusion, we gave preference to national and international consortium guidelines for implementation of pharmacogenomics. We discuss special considerations important in the application of pharmacogenomics to assist clinicians with ordering, interpreting, and applying pharmacogenomics in their practices.

Variation in 100 relevant pharmacogenes among emiratis with insights from understudied populations

Genetic variations have an established impact on the pharmacological response. Investigating this variation resulted in a compilation of variants in "pharmacogenes". The emergence of next-generation sequencing facilitated large-scale pharmacogenomic studies and exhibited the extensive variability of pharmacogenes. Some rare and population-specific variants proved to be actionable, suggesting the significance of population pharmacogenomic research. A profound gap exists in the knowledge of pharmacogenomic variants enriched in some populations, including the United Arab Emirates (UAE). The current study aims to explore the landscape of variations in relevant pharmacogenes among healthy Emiratis. Through the resequencing of 100 pharmacogenes for 100 healthy Emiratis, we identified 1243 variants, of which 63% are rare (minor allele frequency ≤ 0.01), and 30% were unique. Filtering the variants according to Pharmacogenomics Knowledge Base (PharmGKB) annotations identified 27 diplotypes and 26 variants with an evident clinical relevance. Comparison with global data illustrated a significant deviation of allele frequencies in the UAE population. Understudied populations display a distinct allelic architecture and various rare and unique variants. We underscored pharmacogenes with the highest variation frequencies and provided investigators with a list of candidate genes for future studies. Population pharmacogenomic studies are imperative during the pursuit of global pharmacogenomics implementation.

Towards precision medicine: interrogating the human genome to identify drug pathways associated with potentially functional, population-differentiated polymorphisms

Drug response variations amongst different individuals/populations are influenced by several factors including allele frequency differences of single nucleotide polymorphisms (SNPs) that functionally affect drug-response genes. Here, we aim to identify drugs that potentially exhibit population differences in response using SNP data mining and analytics. Ninety-one pairwise-comparisons of >22,000,000 SNPs from the 1000 Genomes Project, across 14 different populations, were performed to identify ‘population-differentiated’ SNPs (pdSNPs). Potentially-functional pdSNPs (pf-pdSNPs) were then selected, mapped into genes, and integrated with drug–gene databases to identify ‘population-differentiated’ drugs enriched with genes carrying pf-pdSNPs. 1191 clinically-approved drugs were found to be significantly enriched (Z > 2.58) with genes carrying SNPs that were differentiated in one or more population-pair comparisons. Thirteen drugs were found to be enriched with such differentiated genes across all 91 population-pairs. Notably, 82% of drugs, which were previously reported in the literature to exhibit population differences in response were also found by this method to contain a significant enrichment of population specific differentiated SNPs. Furthermore, drugs with genetic testing labels, or those suspected to cause adverse reactions, contained a significantly larger number (P < 0.01) of population-pairs with enriched pf-pdSNPs compared with those without these labels. This pioneering effort at harnessing big-data pharmacogenomics to identify ‘population differentiated’ drugs could help to facilitate data-driven decision-making for a more personalized medicine.

Computer aided analysis of disease linked protein networks

Proteins can interact in various ways, ranging from direct physical relationships to indirect interactions in a formation of protein-protein interaction network. Diagnosis of the protein connections is critical to identify various cellular pathways. Today constructing and analyzing the protein interaction network is being developed as a powerful approach to create network pharmacology toward detecting unknown genes and proteins associated with diseases. Discovery drug targets regarding therapeutic decisions are exciting outcomes of studying disease networks. Protein connections may be identified by experimental and recent new computational approaches. Due to difficulties in analyzing in-vivo proteins interactions, many researchers have encouraged improving computational methods to design protein interaction network. In this review, the experimental and computational approaches and also advantages and disadvantages of these methods regarding the identification of new interactions in a molecular mechanism have been reviewed. Systematic analysis of complex biological systems including network pharmacology and disease network has also been discussed in this review.

Deep learning in pharmacogenomics: from gene regulation to patient stratification

This Perspective provides examples of current and future applications of deep learning in pharmacogenomics, including: identification of novel regulatory variants located in noncoding domains of the genome and their function as applied to pharmacoepigenomics; patient stratification from medical records; and the mechanistic prediction of drug response, targets and their interactions. Deep learning encapsulates a family of machine learning algorithms that has transformed many important subfields of artificial intelligence over the last decade, and has demonstrated breakthrough performance improvements on a wide range of tasks in biomedicine. We anticipate that in the future, deep learning will be widely used to predict personalized drug response and optimize medication selection and dosing, using knowledge extracted from large and complex molecular, epidemiological, clinical and demographic datasets.