Single subject transcriptome analysis to identify functionally signed gene set or pathway activity

• Published in: Biocomputing 2018 Vol. 23; pp. 400 - 411
• Main Authors: Berghout, Joanne, Li, Qike, Pouladi, Nima, Li, Jianrong, Lussier, Yves A.
• Format: Book Chapter Journal Article
• Language: English
• Published: United States WORLD SCIENTIFIC 01.01.2018
• Subjects: Animals; Computational Biology - methods; Databases, Genetic - statistics & numerical data; Diet, High-Fat - adverse effects; Female; Gene Expression Profiling - statistics & numerical data; Gene Ontology - statistics & numerical data; Gene Regulatory Networks; Humans; Male; Mice; Mice, 129 Strain; Mice, Inbred C57BL; Mice, Inbred DBA; Mice, Inbred NZB; Mice, Inbred Strains; Oligonucleotide Array Sequence Analysis - statistics & numerical data; Precision Medicine; PRECISION MEDICINE: FROM DIPLOTYPES TO DISPARITIES TOWARDS IMPROVED HEALTH AND THERAPIES; atherosclerosis; N-of-1; reproducibility; transcriptome; high fat; ontology
• ISBN: 9789813235526; 9789813235540; 9813235543; 9813235527; 9789813235533; 9813235535
• ISSN: 2335-6936
• DOI: 10.1142/9789813235533_0037

Analysis of single-subject transcriptome response data is an unmet need of precision medicine, made challenging by the high dimension, dynamic nature and difficulty in extracting meaningful signals from biological or stochastic noise. We have proposed a method for single subject analysis that uses a mixture model for transcript fold-change clustering from isogenically paired samples, followed by integration of these distributions with Gene Ontology Biological Processes (GO-BP) to reduce dimension and identify functional attributes. We then extended these methods to develop functional signing metrics for gene set process regulation by incorporating biological repressor relationships encoded in GO-BP as negatively_regulates edges. Results revealed reproducible and biologically meaningful signals from analysis of a single subject's response, opening the door to future transcriptomic studies where subject and resource availability are currently limiting. We used inbred mouse strains fed different diets to provide isogenic biological replicates, permitting rigorous validation of our method. We compared significant genotype-specific GO-BP term results for overlap and rank order across three replicate pairs per genotype, and cross-methods to reference standards (limma+FET, SAM+FET, and GSEA). All single-subject analytics findings were robust and highly reproducible (median area under the ROC curve=0.96, n=24 genotypes × 3 replicates), providing confidence and validation of this approach for analyses in single subjects. R code is available online at http://www.lussiergroup.org/publications/PathwayActivity