biotmle

biotmle implements statistical methods to identify biomarkers from biological sequencing data (microarrays and RNA-seq) by estimating associations between biomarkers, exposures, and outcomes using Targeted Minimum Loss-Based Estimation (TMLE) and a generalized moderated t-statistic.

Key Features:

• Targeted Minimum Loss-Based Estimation (TMLE): Uses TMLE (van der Laan and Rose 2011) to produce robust, asymptotically linear parameter estimates grounded in causal inference.
• Generalized moderated t-statistic: Incorporates a generalization of Smyth's moderated t-statistic (Smyth 2004) for rigorous statistical evaluation of biomarker associations.
• Confounder adjustment and covariate specification: Adjusts for confounding variables through user-specified covariates when estimating biomarker–outcome associations.
• Supported data types: Applies to microarray and RNA-seq sequencing data for biomarker discovery and differential expression assessment.
• Implementation language: Provided as an R package implementation.

Scientific Applications:

• Biomarker discovery in genomics: Identify candidate biomarkers associated with exposures or outcomes in genomic datasets.
• Differential expression analysis: Assess differential expression signals in microarray and RNA-seq data using moderated statistics and causal estimation.
• Causal effect estimation: Estimate causal parameters relating biomarkers to outcomes in observational studies using TMLE.
• Translational research and personalized medicine: Support biomarker-based investigations relevant to disease mechanisms and therapeutic strategies.

Methodology:

Combines Targeted Minimum Loss-Based Estimation (TMLE) and a generalized moderated t-statistic with adjustment for confounders; implemented as an R package.