Random forest-integrated analysis in AD and LATE brain transcriptome-wide data to identify disease-specific gene expression

Abstract

Alzheimer’s disease (AD) is a complex neurodegenerative disorder that affects thinking, memory, and behavior. Limbic-predominant age-related TDP-43 encephalopathy (LATE) is a recently identified common neurodegenerative disease that mimics the clinical symptoms of AD. The development of drugs to prevent or treat these neurodegenerative diseases has been slow, partly because the genes associated with these diseases are incompletely understood. A notable hindrance from data analysis perspective is that, usually, the clinical samples for patients and controls are highly imbalanced, thus rendering it challenging to apply most existing machine learning algorithms to directly analyze such datasets. Meeting this data analysis challenge is critical, as more specific disease-associated gene identification may enable new insights into underlying disease-driving mechanisms and help find biomarkers and, in turn, improve prospects for effective treatment strategies. In order to detect disease-associated genes based on imbalanced transcriptome-wide data, we proposed an integrated multiple random forests (IMRF) algorithm. IMRF is effective in differentiating putative genes associated with subjects having LATE and/or AD from controls based on transcriptome-wide data, thereby enabling effective discrimination between these samples. Various forms of validations, such as cross-domain verification of our method over other datasets, improved and competitive classification performance by using identified genes, effectiveness of testing data with a classifier that is completely independent from decision trees and random forests, and relationships with prior AD and LATE studies on the genes linked to neurodegeneration, all testify to the effectiveness of IMRF in identifying genes with altered expression in LATE and/or AD. We conclude that IMRF, as an effective feature selection algorithm for imbalanced data, is promising to facilitate the development of new gene biomarkers as well as targets for effective strategies of disease prevention and treatment.

Risk Factors and Prediction Models for Nonalcoholic Fatty Liver Disease Based on Random Forest

“…Viral infection, autoimmune liver disease, oxidative stress, insulin resistance, heredity, and intestinal flora disorder can all lead to this disease [ 10 ]. More and more research is being paid to exploring the influencing factors of NAFLD, building a crowd risk prediction model, identifying high-risk groups, and diagnosing and preventing NAFLD in advance [ 19 ]. Therefore, a good prediction model will accurately predict the progress of the disease, so as to effectively monitor and timely intervene the high-risk groups.…”

“…In recent years, machine learning based on random forest (RF) and other algorithm have been widely applied in the medical field [19][20][21][22][23]. Therefore, this study is aimed at establishing a risk prediction model of NAFLD based on routine physical examination indexes through a retrospective cohort study and at providing new ideas for the early identification of high-risk groups of NAFLD patients.…”

Alcohol Intake Differentiates AD and LATE: A Telltale Lifestyle from Two Large-Scale Datasets

“…We adopted three existing clinical or neuropathological diagnostic criteria to categorize whether a subject had AD and/or LATE: 1) Braak score [19, 20]; 2) CERAD score [20, 21]; 3)TDP-43 stage [22]. We followed the detailed rules in [23] for categorization. Taking ROSMAP as an example, we used the first two scores for annotating subjects with AD and the third measure for LATE.…”

“…The ROSMAP data used in our analysis was substantially imbalanced in terms of the sample sizes of controls and cases in many strata. In the literature, ML algorithms were developed to meet the challenges of imbalanced data for classification [24, 25] and for FS [23]. However, this ML-based study needed to consider selecting features to account for feature representativeness and inter-correlations and classifying examples jointly.…”

“…If a part of the samples from the majority class(es) had been used, it would have under-utilized the clinically valuable data and led to misleading importance for the variables. On the other hand, as pointed out in [23], if the majority class(es) had been bootstrapped many times while the remaining minority class(es) had been kept the same, then it would have hardly ensured a sufficient number of used sample subsets for the feature ranking model and, consequently, it would have led to degraded learning performance and made the selected features unstable. To overcome the challenges of imbalanced data, we developed Algorithm 1: First, at each iteration, we generated a new subset by concatenating the minority samples with the under-sampled majority samples, and we applied the FS algorithm to this subset to score the features; next, we averaged the scores obtained from all implemented iterations to get the ranking of all the features.…”