A growing number of long non-coding RNAs (lncRNAs) have emerged as vital metabolic regulators. However, most human lncRNAs are non-conserved and highly tissue-specific, vastly limiting our ability to identify human lncRNA metabolic regulators (hLMRs). In this study, we establish a pipeline to identify putative hLMRs that are metabolically sensitive, disease-relevant, and population applicable. We first progressively processed multilevel human transcriptome data to select liver lncRNAs that exhibit highly dynamic expression in the general population, show differential expression in a nonalcoholic fatty liver disease (NAFLD) population, and response to dietary intervention in a small NAFLD cohort. We then experimentally demonstrated the responsiveness of selected hepatic lncRNAs to defined metabolic milieus in a liver-specific humanized mouse model. Furthermore, by extracting a concise list of protein-coding genes that are persistently correlated with lncRNAs in general and NAFLD populations, we predicted the specific function for each hLMR. Using gain-and loss-of-function approaches in humanized mice as well as ectopic expression in conventional mice, we validated the regulatory role of one nonconserved hLMR in cholesterol metabolism by coordinating with an RNA-binding protein, PTBP1, to modulate the transcription of cholesterol synthesis genes. Our work overcome the heterogeneity intrinsic to human data to enable the efficient identification and functional definition of diseaserelevant human lncRNAs in metabolic homeostasis.