Epigenetic clocks, DNA methylation-based biomarkers, accurately measure age withinspecific species and tissues but face challenges in multi-species and non-mammalian contexts. This study aims to enhance the construction of biologically meaningful, multi-species epigenetic clocks, particularly applicable to both humans and animals. Utilizing supervised machine learning on DNA methylation data, this research incorporates biological information, such as genome mapping and life history traits, to improve clock accuracy and interpretability. Findings include the development of diverse epigenetic clocks for various mammalian species, enabled by robust statistical methods and a reproducible software pipeline. An R package, MammalMethylClock, was introduced to facilitate the construction, assessment, and application of these clocks, incorporating existing models from the Mammalian Methylation Consortium. This package supports translational biomedical research by enhancing the study of age-related chronic diseases across different model organisms. Moreover, the remainder of this study examines the research problem of accelerated aging in people living with HIV (PLWH) and the mitigating effects of highly active antiretroviral therapy (HAART). Previous research indicates that HIV-related age acceleration begins shortly after seroconversion. This longitudinal study within the Multicenter AIDS Cohort Study (MACS) examines 60 PLWH and 60 matched controls across four specific time points: pre- seroconversion, closest to seroconversion, pre-HAART, and post-1-3 years of HAART. The study uses comprehensive epigenetic clock data and flow cytometry measures, including naive, senescent, activated, and total CD4/CD8 counts. The method involves analyzing epigenetic aging through five different aging clocks at these time points to assess the impact of HIV on aging and the partial restoration effects of HAART on immune function and aging patterns. By tracking these changes, the study aims to identify specific aging patterns associated with HIV and guide potential interventions. The findings reveal that HIV infection significantly accelerates epigenetic aging in most clocks, and this acceleration is only partially mitigated by HAART. This suggests that HIV contributes to the premature onset of clinical aging through its profound effects on the immune system, even when viral replication is controlled. The study underscores the need for interventions to enhance the healthspan of PLWH and offers insights into the broader aging process.

In a variety of settings, including the medical field, it is common for the number of variables gathered to far exceed the sample size. Along with a high dimension, many of these included variables are often correlated. This can pose problems for traditional methods. Much of the time, the data cannot be utilized completely as is, but instead requires previous research to guide researchers to choose relevant predictors prior to model selection. Traditional methods such as logistic regression and mixed models cannot necessarily converge and struggle with identifiability when the number of measurements collected approach or become larger than the number of patients in the study. Machine-learning techniques, including Random Forests and the newly developed Fuzzy Forests method, can accommodate data with high dimensionality. We concentrate on decision trees in particular because of their relative ease of use, availability and predictive ability. Random Forest is a widely used, parallelizable and computationally efficient method; however it does not acknowledge any correlation between variables leading to a preference for correlated predictors. Fuzzy Forest, on the other hand, explicitly explores the correlation structure among the variables, leading to unbiased variable importance measures. Fuzzy Forest, along with Random Forest, is utilized in three applications; smoking cessation in health care workers, re-arrest among homeless ex-offenders and genetic predictors of lithium response in individuals with Bipolar disorder.

In this paper we introduce fuzzy forests, a novel machine learning algorithm for ranking

the importance of features in high-dimensional classication and regression problems.

Fuzzy forests is specically designed to provide relatively unbiased rankings of variable

importance in the presence of highly correlated features, especially when p >> n . We

introduce our implementation of fuzzy forests in the R  package, fuzzyforest . Fuzzy forests

works by taking advantage of the network structure between features. First, the features

are partitioned into separate modules such that the correlation within modules is high

and the correlation between modules is low. The package fuzzyforest  allows for easy use

of Weighted Gene Coexpression Network Analysis (WGCNA) to form modules of features

such that the modules are roughly uncorrelated. Then recursive feature elimination random

forests (RFE-RFs) are used on each module, separately. From the surviving features,

a nal group is selected and ranked using one last round of RFE-RFs. This procedure

results in a ranked variable importance list whose size is pre-specied by the user. The

selected features can then be used to construct a predictive model.

When the number of covariates is small, nonparametric regression methods serve a number of useful purposes. In this setting, nonparametric regression methods often demonstrate better predictive performance than parametric models. This is because nonparametric methods have the virtue of being able to detect nonlinear structure and complex interactions. In settings where the sample size is small or the level of noise is high, it may be the case that parametric models outperform nonparametric methods. However, even in this setting, nonparametric methods can be useful for diagnosing problems of model misspecification. Unfortunately, when the number of covariates is large, the curse of dimensionality, in its many forms, renders many of the most commonly used nonparametric regression meth- ods unstable and prone to overfitting. We have developed two methods that, in some sense, overcome the curse of dimensionality. Both methods implicitly assume the existence of lower dimensional structure. First, we have developed a variant of random forests, called fuzzy forests. Fuzzy forests reduce the bias observed in random forest variable importance measures by clustering covariates into distinct groups such that the correlation of covariates within a group is high and the cor- relation between groups is low. Fuzzy forests is expected to work well when the true regression function exhibits an additive structure. Second, we have extended a machine learning method called metric learning to right-censored survival out- comes. If the true regression function is multi-index, we have shown that a closely related metric learning estimator achieves a rate of convergence dependent on the number of indices rather than the number of covariates.

Background

Objective

Methods

Results

Conclusions