RNAseq (lncRNA) -materials and methods

RNA Extraction Preparation for Next-Generation SequencingTotal RNA was extracted using TRK-1001 total RNA purification kit (LC Science, Houston, TX) following the manufacturer's procedure. The total RNA quantity and purity were analysis of Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA) with RIN number >7.0. Approximately 5 ug of total RNA representing a specific adipose type was used to deplete ribosomal RNA according to the manufacturer’s instructions of the Ribo-Zero™ rRNA Removal kit (Epidemiology version) (Epicentre, an Illumina company, Madison, WI). Following purification, the poly(A)- or poly(A)+ RNA fractions is fragmented into small pieces using divalent cations under elevated temperature. Then the cleaved RNA fragments were reverse-transcribed to construct cDNA library using the dUTP method as described [1], the average insert size for the paired-end libraries was 300 bp (±50 bp). RNA libraries were then sequenced on the Illumina HiSeq 2000 or 2500 platform using 125 bp paired-end reads.RNA-Sequencing Data AnalysisAll RNA-seq data were aligned to hg19 using TopHat[2]v2.0.9 with default parameters. The mapped reads were assembled using Cufflinks[3] v2.11. All multiple assembled transcript files (GTF format) were then merged to produce a unique set of transcriptomes using the Cuffmerge utility provided by the Cufflinks package. Cuffdiff v2.11 was used for all differential expression analyses with Gencode v21.0 annotation. Transcript abundances were estimated by Cufflinks in Fragments Per Kilobase per Million mapped reads (FPKM) for pairedend reads or Reads Per Kilobase per Million mapped reads (RPKM) for single-end reads[4]. In all differentialexpression tests, a gene was considered significant if the q value was less than 0.05 (Cuffdiff Default).lncRNA DiscoveryWe filtered the assembled novel transcripts 2 cell lines to obtain putative lncRNAs. Firstly, identical and overlapping transcripts were merged to remove redundancy. Then, transcripts overlapping with known exons of genes were removed. Only transcripts with length>200nt were retained. To identify potential known lncRNA transcripts, we compare the merged transcriptome with lncRNA and protein-coding genes in public databases including GENCODE[5], NONCODE[6], RefSeq [7], and Ensembl[8]to eliminate potential known lncRNA transcripts with further filtering of length>200 nucleotides.In order to obtain a reliable dataset of putative lncRNAs, single exon models were filtered out. Next, we removed transcripts that were likely to be assembly artifacts or PCR run-on fragments according to class code annotated by Cuffmerage. Among the different classes, only those a nnotated by ‘‘u’’, ‘‘i’’and ‘‘x’’ were retained, which repres ent novel intergenic, intronic and cis-antisense transcripts, respectively. But here, most analyses were focused on intergenic, intronic and cis antisense lncRNAs. Extremely low gene expression is generally considered to be transcriptional noise [9]. On average, 85% of the initial reads could be aligned to the hg19 assembly of the human genome sequence. Transcripts with RPKM/FPKM under lower bound of single tail 85% confidence interval (<0.3) for all expression values were removed.Lastly, we calculated the protein-coding capacity of novel transcripts using CPC (:) which incorporates the sequence features into a support vectormachine to assess the protein-coding potential of each transcript. The proportion of coding transcripts miss-classified as non-coding RNAs by CPC was previously shown to be marginal [10,11], suggesting CPC is a robust approach for distinguishing coding from noncoding RNAs. Then those putative transcripts with CPC score<-1 were retained as candidate lncRNAs for the further analysis.Putative lncRNA ClassificationThe assembled putative lncRNAs were divided into three categories: (1) lncRNAs without any overlaps with any genes (RefSeq or Ensembl) were classified as intergenic overlap lncRNAs (intergenic lncRNAs); (2) lncRNAs that were entirely contained within intron of any protein-coding genes in either sense or antisense orientation were classified as intronic overlap lncRNAs (intronic lncRNAs); (3) lncRNAs with exonic overlaps with any exons of RefSeq transcript on the opposite strand were classified as cis-antisense overlap lncRNAs (cis-antisense lncRNAs).Gene Function Enrichment AnalysisDAVID (:/) [14] was used to perform gene function enrichment analysis based on GO (:/) [12], KEGG (:genome.jp/kegg/) [13] annotation by submitting closest gene lists for putative intergenic lncRNAs, host genes for putative embryonic brain intronic lncRNAs and overlapping genes for putative embryonic brain cis-antisense lncRNAs, respectively. Only putative intergenic lncRNAs with distance to closest genes <500 kb were kept for this analysis, which would eliminate longdistance irrelevant genes. Furthermore, putative intronic lncRNAs embedded in long introns (<100 kb) of known genes were also discarded to avoid the bias of large introns. Functional terms with Benjamini-Hochberg adjusted p-values<0.05 were considered to be significantly。