Search bioRxiv⌕ Search

Biology subjects

Maciejewski, M.

Publications and source records attributed to Maciejewski, M..

2 recordsLinked to original sources
bioRxiv2022

Meta-Analyzed Atopic Dermatitis Transcriptome (MAADT) defines a strong correlation to disease activity and consistent to therapeutic effect

BackgroundAtopic Dermatitis (AD) is a persistent inflammatory disease of the skin to which a few novel treatment options have recently become available. Multiple published datasets, from RNA sequencing (RNA-seq) and microarray experiments performed on lesional (LS) and non-lesional (NL) skin biopsies collected from AD patients, provide a useful resource to better define an AD gene signature and evaluate therapeutic effects. MethodsWe evaluated 22 datasets using defined selection criteria and leave-one-out analysis and then carried out a meta-analysis (M-A) to combine 4 RNA-seq datasets and 5 microarray datasets to define a disease gene signature for AD skin tissue. We used this gene signature to evaluate its correlation to disease activity in published AD datasets, as well as the treatment effect of some of the existing and experimental therapies. ResultsWe report the AD gene signatures developed separately from the RNA-seq or the microarray datasets, as well as a gene signature from datasets combined across these two technologies; all 3 gene signatures showed a strong correlation to the disease activity score (SCORAD) - microarray: Pearsons{rho} = 0.651, p-value < 0.01, RNA-seq:{rho} = 0.640, p-value < 0.01, combined:{rho} = 0.649, p-value < 0.01. The gene signature improvement (GSI) of two existing effective therapies, Dupilumab and Cyclosporine, as well as that of other experimental treatments, is consistent with their reported cohort level efficacy from the associated clinical trials. ConclusionsThe M-A derived AD gene signature provides an evolution of an important resource to correlate gene expression to disease activity and will be helpful for evaluating potential treatment effects for novel therapies.

bioinformatics↗
bioRxiv2019

Deep Learning of Representations for Transcriptomics-based Phenotype Prediction

The ability to predict health outcomes from gene expression would catalyze a revolution in molecular diagnostics. This task is complicated because expression data are high dimensional whereas each experiment is usually small (e.g.,[~] 20,000 genes may be measured for[~] 100 subjects). However, thousands of transcriptomics experiments with hundreds of thousands of samples are available in public repositories. Can representation learning techniques leverage these public data to improve predictive performance on other tasks? Here, we report a comprehensive analysis using different gene sets, normalization schemes, and machine learning methods on a set of 24 binary and multiclass prediction problems and 26 survival analysis tasks. Methods that combine large numbers of genes outperformed single gene methods, but neither unsupervised nor semi-supervised representation learning techniques yielded consistent improvements in out-of-sample performance across datasets. Our findings suggest that using l2-regularized regression methods applied to centered log-ratio transformed transcript abundances provide the best predictive analyses.

bioinformatics↗