FANTOM CAT lncRNAome

From Wiki
Jump to navigationJump to search


Overview: What is FANTOM CAT?

The human genome is pervasively transcribed, generating numerous long non-coding RNAs (lncRNA). As the majority of these lncRNAs are lowly expressed and relatively short, it is unclear how many of the observed lncRNAs are results of spurious transcription events. Also, due to their low expression levels, building a reliable and complete catalogue of lncRNAs is challenging. To understand the origin and expression landscape of lncRNA, we sought to create a comprehensive and well-annotated human transcriptome, which is well-supported with transcription initiation evidences and linked to a large number of expression datasets.

Introducing the FANTOM5 CAGE Associated Transcriptome, i.e. FANTOM CAT, a meta-assembly of human transcriptome, based on association of FANTOM5 CAGE clusters with non-redundant transcript models merged from GENCODE, FANTOM, ENCODE, Human Body Map and miTranscriptome.

Main features of FANTOM CAT include,

  1. Broad Scope: GENCODE19 + ~7600 RNASEQ libraries + ~1900 CAGE libraries
  2. Transcription Initiation Evidence: All transcripts are supported by FANTOM5 CAGE clusters, with 1-to-1 promoter-to-transcript association, quantified by Transcription Initiation Evidence Score (TIEScore).
  3. Gene Class Annotation: All genes were categorized into gene classes based on existing annotation, coding potential, loci architecture, promoter directionality, epigenomic feature etc.
  4. Wide Expression Landscape: All genes were linked to expression profile from 1897 CAGE libs on primary cells, tissues, cell lines, differentiation/activation time courses + ~100 ENCODE CAGE libs on cell fractions, polyA+/- etc.


Data Files

Documentaions and data files of the FANTOM CAT can be found on webdav


Paper Outline and Tasks (updating)

Section 1: Building CAGE Associated Transcriptome (CAT) for lncRNA catalogue 

Rationale: Due to the genrally lower expression level of lncRNAs and background noisy of RNASeq assemblies, building a reliable and comphrehensive lncRNA catalogue is challenging. Introducing FANTOM CAT, a meta-assembly of human transcriptome, based on association of FANTOM5 CAGE DPI clusters with non-redundant transcript models merged from GENCODE, FANTOM, ENCODE, Human Body Map and miTranscriptome. FANTOM CAT was built with an emphasis on transcription initation evidence. To evaluate the association between transcripts and FANTOM5 CAGE DPI clusters, we used Transcription Initiation Evidence Score, i.e. TIEScore. TIEScore is an empirical measurement on the evidence of transcription initiation, based on transcript length, exon number, CAGE tag count and distance between transcript 5'end & CAGE cluster. Every transcript in FANTOM CAT is associated with ONE FANTOM5 CAGE DPI cluster, based on TIEScore. We settled on 3 cutoffs, i.e. permissive, robust and stringent, representing 50%, 96.8% and 100% specificity respectively. Most of the analyses should be done on the robust subset. For documentation and data files of FANTOM CAT, visit webdav.


1. Meta-assembly of FANTOM CAT using TIEScore

  • Idea: Pool diverse transcript soruces to maximize the scope; Associate to CAGE using TIEScore to maximize robustness;
  • People: Chung-Chau Hon, Jordan Ramilowski
  • Task: 
  • Ref:


2. Benchmarks against GENCODE & other lncRNA catalogues

  • Idea: To demonstrate FANTOM CAT is more robust than previous lncRNA catalog in terms of lncRNA TSS recovery
  • People: Chung-Chau Hon
  • Task:
  • Ref:


Section 2: Genomic Origins of lncRNA: Promoter vs Enhancer-derived

Rationale:Depends on different studies, the number of lncRNA "genes" in human genome is estimated around ~10,000 to ~50,000. Although a handful of these lncRNA genes were shown to share similar properties with coding genes, e.g. K4-K36 domain, vast majority of lncRNAs are lowly expressed, relatively short and relatively more tissue specific. It implies the vast majority of lncRNAs are likely to be produced / destroyed in ways that are substantially different from that of coding mRNAs. For instance, both promoter and ehancers are being transcribed, it is currently unclear that how much of the catalogued lncRNA were derived from promoters versus enhancers. A previous study in mouse suggests "chromatin signatures at transcriptional start sites separate two equally populated yet distinct classes of intergenic long noncoding RNAs", i.e. promoter-derived and enhancer-derived lncRNA. Also a recently study of ROADMAP histone marks suggest a larger fraction of lincRNA TSSs is marked in a tissue-specific manner by H3K4me1 than by H3K4me3, implying majority of annotated lncRNAs are more likely be derived from enhancer-like regions. Here we sought to compare the prevalence and properties promoter-derived and enhancer-derived lncRNAs, by intersecting FANTOM CAT lncRNA with various external datasets, e.g. ROADMAP epigenome datasets, and compare the properties of p_lncRNA vs e_lncRNA in general.


1. Prevalence of Promoter vs Enhancer-derived lncRNAs

  • Idea: By intersecting with ROADMAP DHS regions, we could classify lncRNA as promoter-derived and enhancer-derived lncRNA. Preliminary data suggest vast marjority of lncRNA in FANTOM CAT are enhancer-derived.
  • People: Chung-Chau Hon
  • Task:
  • Ref: 


2. Epigenomic origin of lncRNA TSS


3. Motifs: Core promoter conservation and TFBS


4. Stability, U1/PAS sites and TSS directionality


5. Transposable elements derived lncRNAs


6. Evolutionary Conservation of lncRNAs

  • Idea: Conservation of p_lncRNA and e_lncRNA between human and mouse. Conserved co-expression network?
  • People: Are you guys still interested ?? ---> Cedric Notredame
  • Task:
  • Ref:


Section 3: Inference of Potentially Functional p_lncRNAs and e_lncRNAs

Rationale:  Every gene in FANTOM CAT is linked to over 2000 expression profiles on primary cells, tissues, cell lines, differentiation/activation time courses, cell fractionation etc. The source data include all FANTOM5 CAGE, ENCODE CAGE and Fort's Stem Cell CAGE. By exploring the expression landscape of lncRNAs, we sought to identify potentially functional lncRNA, through mainly differetial expression analyses and expression correlations among samples.


1. By Differential expression: Specifically Paired Samples

  • Idea: Identify differenetially expressed p_lncRNA and e_lncRNA in specifically paired samples, e.g. cancer VS non-cancer in phase 1 dataset, data points in time courses phase 2 dataset, nuclear vs cytosol in ENCODE datasets, iPSC transformation in Fort's dataset etc. 
  • People: Bogumil Kaczkowski, Chung-Chau Hon
  • Task:
  • Ref:


2. By Expression correlation A: Guilt-by-Association using GSEA

  • Idea: Calculate global correlation between lncRNA and mRNA, then rank the mRNA by correlation, and look for enrichment of functional by GSEA prerank, like this.
  • People: Sebastian Schmeier, are you guys still interested? ---> Owen Rackham, Eivind Valen
  • Task:
  • Ref:


3. By Expression correlation B: Correlated Gene Pairs, e.g. mRNA <-> e_lncRNA

  • Idea: To identify significantly correlation gene pairs in different scenerios, e.g. antisense lncRNA vs mRNA, e_lncRNA vs neighbor mRNA etc. Predefined host-guest gene pairs can be found on webdav, a good point to start.
  • People: Chung-Chau Hon, anyone else? maybe Makis?
  • Task:
  • Ref:


4. By Temporal Shift Analyses in Time Course datasets


5. By Tissue specificity and Sample Ontology Enrichment Analyses


6. By Disease GWAS association

  • Idea: To overlay Disease GWAS association onto FANTOM CAT lncRNA and intersect with the above functional evidences
  • People: Sebastian Schmeier
  • Task:
  • Ref:
Retrieved from "http://fantom5-collaboration.gsc.riken.jp/wiki/index.php?title=FANTOM_CAT_lncRNAome&oldid=7934"