A comparison of RNA-Seq data preprocessing pipelines for transcriptomic predictions across independent studies.

BACKGROUND: RNA sequencing combined with machine learning techniques has provided a modern approach to the molecular classification of cancer. Class predictors, reflecting the disease class, can be constructed for known tissue types using the gene expression measurements extracted from cancer patients. One challenge of current cancer predictors is that they often have suboptimal performance estimates when integrating molecular datasets generated from different labs. Often, the quality of the data is variable, procured differently, and contains unwanted noise hampering the ability of a predictive model to extract useful information. Data preprocessing methods can be applied in attempts to reduce these systematic variations and harmonize the datasets before they are used to build a machine learning model for resolving tissue of origins. RESULTS: We aimed to investigate the impact of data preprocessing steps-focusing on normalization, batch effect correction, and data scaling-through trial and comparison. Our goal was to improve the cross-study predictions of tissue of origin for common cancers on large-scale RNA-Seq datasets derived from thousands of patients and over a dozen tumor types. The results showed that the choice of data preprocessing operations affected the performance of the associated classifier models constructed for tissue of origin predictions in cancer. CONCLUSION: By using TCGA as a training set and applying data preprocessing methods, we demonstrated that batch effect correction improved performance measured by weighted F1-score in resolving tissue of origin against an independent GTEx test dataset. On the other hand, the use of data preprocessing operations worsened classification performance when the independent test dataset was aggregated from separate studies in ICGC and GEO. Therefore, based on our findings with these publicly available large-scale RNA-Seq datasets, the application of data preprocessing techniques to a machine learning pipeline is not always appropriate.

Correction to: Transcriptome analyses of tumor-adjacent somatic tissues reveal genes co-expressed with transposable elements.

[This corrects the article DOI: 10.1186/s13100-019-0180-5.].

Transcriptome analyses of tumor-adjacent somatic tissues reveal genes co-expressed with transposable elements.

BACKGROUND: Despite the long-held assumption that transposons are normally only expressed in the germ-line, recent evidence shows that transcripts of transposable element (TE) sequences are frequently found in the somatic cells. However, the extent of variation in TE transcript levels across different tissues and different individuals are unknown, and the co-expression between TEs and host gene mRNAs have not been examined. RESULTS: Here we report the variation in TE derived transcript levels across tissues and between individuals observed in the non-tumorous tissues collected for The Cancer Genome Atlas. We found core TE co-expression modules consisting mainly of transposons, showing correlated expression across broad classes of TEs. Despite this co-expression within tissues, there are individual TE loci that exhibit tissue-specific expression patterns, when compared across tissues. The core TE modules were negatively correlated with other gene modules that consisted of immune response genes in interferon signaling. KRAB Zinc Finger Proteins (KZFPs) were over-represented gene members of the TE modules, showing positive correlation across multiple tissues. But we did not find overlap between TE-KZFP pairs that are co-expressed and TE-KZFP pairs that are bound in published ChIP-seq studies. CONCLUSIONS: We find unexpected variation in TE derived transcripts, within and across non-tumorous tissues. We describe a broad view of the RNA state for non-tumorous tissues exhibiting higher level of TE transcripts. Tissues with higher level of TE transcripts have a broad range of TEs co-expressed, with high expression of a large number of KZFPs, and lower RNA levels of immune genes.

Developmental decoupling of alternative phenotypes: insights from the transcriptomes of horn-polyphenic beetles.

Developmental mechanisms play an important role in determining the costs, limits, and evolutionary consequences of phenotypic plasticity. One issue central to these claims is the hypothesis of developmental decoupling, where alternate morphs result from evolutionarily independent developmental pathways. We address this assumption through a microarray study that tests whether differences in gene expression between alternate morphs are as divergent as those between sexes, a classic example of developmental decoupling. We then examine whether genes with morph-biased expression are less conserved than genes with shared expression between morphs, as predicted if developmental decoupling relaxes pleiotropic constraints on divergence. We focus on the developing horns and brains of two species of horned beetles with impressive sexual- and morph-dimorphism in the expression of horns and fighting behavior. We find that patterns of gene expression were as divergent between morphs as they were between sexes. However, overall patterns of gene expression were also highly correlated across morphs and sexes. Morph-biased genes were more evolutionarily divergent, suggesting a role of relaxed pleiotropic constraints or relaxed selection. Together these results suggest that alternate morphs are to some extent developmentally decoupled, and that this decoupling has significant evolutionary consequences. However, alternative morphs may not be as developmentally decoupled as sometimes assumed and such hypotheses of development should be revisited and refined.