Long intergenic non-coding RNA detection benefited from integrative modeling of (Epi)genomic data

Prediction of long intergenic non-coding RNAs (lincRNAs) is a prerequisite to analyze sequence features of non-coding RNAs and explore their regulatory function. Genomic sequence features provide fundamental backgrounds for lincRNA predictions, due to that sequence information at least partially aids such predictions. However, genomic sequence alone seems to reach an end involving sensitivity for lincRNA prediction in eukaryotes. Chromatin factors leave marks that can be captured by high-throughput approaches such as ChIP-seq are also important features, as revealed by previous studies. We demonstrate that the performance of lincRNA predictions can be improved when incorporating both high-throughput chromatin modification and genomic sequence features by logistic regression with LASSO regularization. The discriminating features include H3K4me1, H3K27ac, H3K9me3, Open Reading Frames and several repeat elements. Importantly, chromatin information is suggested to be complementary to genomic sequence information, highlighting the importance of an integrated model. We also show that the lincRNA expression specificity can be efficiently modeled by the chromatin data with same developmental stage. The study not only supports the biological hypothesis that chromatin factors can regulate developmental-stage-specific expression of lincRNAs, also reveals the discriminating features between lincRNA and coding genes.