Integrated omics technologies have been instrumental in a host of applications including disease discovery. In ESR 12, the main objectives include the identification of the molecular signatures (using multi-omics analysis of transcriptomic and metabolic data) of co-exposure to neurotoxicants in the human biosamples. Multi-omics are a powerful tool in disease discovery that have undergone large advancements in recent years; available technologies allow for the generation of datasets with tens of thousands of outputs. This requires large computational power and advanced statistical analysis for datasets. While the in-house methods for metabolomics analysis are well-established, there is a need for a statistical pipeline for transcriptomic data analysis. Unfortunately, there is debate within the scientific community over the ideal methods to process raw data, analyze and identify differentially expressed genes (DEGs), and generate robust results while limiting Type I and Type II errors.
We developed two statistical method pipelines using Linear Models for Microarray Data (LIMMA; R software) and moderated t-test (Agilent GeneSpring™ software) for DEG discovery and functionality comparisons. R is an open-source software that allows extensive control of statistical pipelines and parameters and the ability to handle complex statistical designs, although it requires basic knowledge of programming languages. GeneSpring™ is a commercially available, user-friendly analysis software that is efficient for basic statistical analyses with the benefit of freely available technical support. However, Genespring™ is geared towards parametric data analysis and tends to be rigid in its pipeline capabilities.
We analyzed Agilent™ microarray data generated from three real datasets from experiments from within the lab group (One-Color, SurePrint Zebrafish Gene Expression v3 4 x 44k Microarray, design ID: 026437) that aim to detect the molecular mechanisms involved in metabolic disorders associated with environmental contaminants. 3-day post fertilization (dpf) zebrafish larvae (n = ~17 per replicate, 4 replicates) were exposed for 48 hours (sampled at 5-dpf) to two concentrations of the plasticizer bis(2-ethylhexyl) phthalate (DEHP; 25 nM and10 µM), positive control (amiodarone; 1 µM), and a carrier control [dimethyl sulfoxide (DMSO); 0.1%]. Samples were processed following manufacturer protocols and features were extracted using Agilent Feature Extraction™ software. Raw data was exported and analyzed using the above methods.
Preliminary results suggest that LIMMA is more conservative than the moderated t-test and generates fewer DEGs. Overall, this work is instrumental in future efforts to generate statistically robust and reliable computational models and systems biology for the prediction of a host of metabolic diseases.