binomialRF: interpretable combinatoric efficiency of random forests to identify biomarker interactions.

BACKGROUND: In this era of data science-driven bioinformatics, machine learning research has focused on feature selection as users want more interpretation and post-hoc analyses for biomarker detection. However, when there are more features (i.e., transcripts) than samples (i.e., mice or human samples) in a study, it poses major statistical challenges in biomarker detection tasks as traditional statistical techniques are underpowered in high dimension. Second and third order interactions of these features pose a substantial combinatoric dimensional challenge. In computational biology, random forest (RF) classifiers are widely used due to their flexibility, powerful performance, their ability to rank features, and their robustness to the "P > > N" high-dimensional limitation that many matrix regression algorithms face. We propose binomialRF, a feature selection technique in RFs that provides an alternative interpretation for features using a correlated binomial distribution and scales efficiently to analyze multiway interactions. RESULTS: In both simulations and validation studies using datasets from the TCGA and UCI repositories, binomialRF showed computational gains (up to 5 to 300 times faster) while maintaining competitive variable precision and recall in identifying biomarkers' main effects and interactions. In two clinical studies, the binomialRF algorithm prioritizes previously-published relevant pathological molecular mechanisms (features) with high classification precision and recall using features alone, as well as with their statistical interactions alone. CONCLUSION: binomialRF extends upon previous methods for identifying interpretable features in RFs and brings them together under a correlated binomial distribution to create an efficient hypothesis testing algorithm that identifies biomarkers' main effects and interactions. Preliminary results in simulations demonstrate computational gains while retaining competitive model selection and classification accuracies. Future work will extend this framework to incorporate ontologies that provide pathway-level feature selection from gene expression input data.

Correction to: binomialRF: interpretable combinatoric efficiency of random forests to identify biomarker interactions.

An amendment to this paper has been published and can be accessed via the original article.

A Community-Based Survey to Assess Knowledge, Attitudes, Beliefs and Practices Regarding Herpes Zoster in an Urban Setting.

INTRODUCTION: In the USA, nearly one in three people will experience herpes zoster (HZ) in their lifetime. Underserved communities may be at even higher risk due to several factors, including access to healthcare, education, and co-morbid conditions. The purpose of this study was to investigate current knowledge, attitudes, beliefs and practices (KABP) relative to HZ and HZ vaccines in a large urban city. METHODS: A cross-sectional KABP survey was conducted via in-person interview among 381 participants aged ≥ 50 years in Detroit, MI, USA, from June to August 2018. Survey results were stratified into two groups [< 60 and ≥ 60 years of age (YO)] for comparison. RESULTS: Of the 381 participants, 373 reported their age (110 < 60 YO and 263 ≥ 60 YO). Overall, the majority of participants reported having heard of HZ and HZ vaccines. In addition, receiving a recommendation from a healthcare provider (37.5%) followed by gaining a better understanding of HZ vaccine (36.7%) and of HZ (29.9%) were leading factors that influenced participants' willingness to receive the vaccine. Of note, 65.5% of participants < 60 YO reported the belief that HZ is preventable versus only 53.2% in those ≥ 60 YO (p = 0.001). CONCLUSION: Our findings underscore the need to educate patients in underserved communities about HZ as well as new HZ vaccine recommendations to improve vaccination rates and reduce the incidence of HZ and its associated sequelae.

Novel disease syndromes unveiled by integrative multiscale network analysis of diseases sharing molecular effectors and comorbidities.

BACKGROUND: Forty-two percent of patients experience disease comorbidity, contributing substantially to mortality rates and increased healthcare costs. Yet, the possibility of underlying shared mechanisms for diseases remains not well established, and few studies have confirmed their molecular predictions with clinical datasets. METHODS: In this work, we integrated genome-wide association study (GWAS) associating diseases and single nucleotide polymorphisms (SNPs) with transcript regulatory activity from expression quantitative trait loci (eQTL). This allowed novel mechanistic insights for noncoding and intergenic regions. We then analyzed pairs of SNPs across diseases to identify shared molecular effectors robust to multiple test correction (False Discovery Rate FDReRNA < 0.05). We hypothesized that disease pairs found to be molecularly convergent would also be significantly overrepresented among comorbidities in clinical datasets. To assess our hypothesis, we used clinical claims datasets from the Healthcare Cost and Utilization Project (HCUP) and calculated significant disease comorbidities (FDRcomorbidity < 0.05). We finally verified if disease pairs resulting molecularly convergent were also statistically comorbid more than by chance using the Fisher's Exact Test. RESULTS: Our approach integrates: (i) 6175 SNPs associated with 238 diseases from ~ 1000 GWAS, (ii) eQTL associations from 19 tissues, and (iii) claims data for 35 million patients from HCUP. Logistic regression (controlled for age, gender, and race) identified comorbidities in HCUP, while enrichment analyses identified cis- and trans-eQTL downstream effectors of GWAS-identified variants. Among ~ 16,000 combinations of diseases, 398 disease-pairs were prioritized by both convergent eQTL-genetics (RNA overlap enrichment, FDReRNA < 0.05) and clinical comorbidities (OR > 1.5, FDRcomorbidity < 0.05). Case studies of comorbidities illustrate specific convergent noncoding regulatory elements. An intergenic architecture of disease comorbidity was unveiled due to GWAS and eQTL-derived convergent mechanisms between distinct diseases being overrepresented among observed comorbidities in clinical datasets (OR = 8.6, p-value = 6.4 × 10- 5 FET). CONCLUSIONS: These comorbid diseases with convergent eQTL genetic mechanisms suggest clinical syndromes. While it took over a decade to confirm the genetic underpinning of the metabolic syndrome, this study is likely highlighting hundreds of new ones. Further, this knowledge may improve the clinical management of comorbidities with precision and shed light on novel approaches of drug repositioning or SNP-guided precision molecular therapy inclusive of intergenic risks.