Integrative omics analysis Qi Liu Center for Quantitative Sciences Vanderbilt University School of Medicine
Content Introduction Data Sources Methods Tools Things to be aware
Why?
Genomics WGS, WES Transcriptomics RNA-Seq Epigenomics Bisulfite-Seq ChIP-Seq Small indels point mutation Copy number variation Structural variation Differential expression Gene fusion Alternative splicing RNA editing Methylation Histone modification Transcription Factor binding Functional effect of mutation Network and pathway analysis Integrative analysis Further understanding of cancer and clinical applications TechnologiesData AnalysisIntegration and interpretationPatient What? at least two different types of omics data
Objectives 1.Understand relationships between different types of molecular data 2.Understand the phenotype – latent: disease subtype – Observable: patient outcome
Data sources TCGA
Firehose
cBioPortal
ICGC
COSMIC
ENCODE
FANTOM
GTEX
Methods Sequential or overlap analysis Clustering Correlation analysis Linear regression Network based analysis Bayesian …..
Sequential or overlap analysis Confirmation or refinement of findings – Each data are independently analyzed to get a list of interesting entities – Lists of interesting entities are linked together Chin, K. et al. Genomic and transcriptional aberrations linked to breast cancer pathophysiologies. Cancer Cell 10, 529–541 (2006). Lando, M. et al. Gene dosage, expression, and ontology analysis identifies driver genes in the carcinogenesis and chemoradioresistance of cervical cancer. PLoS Genet. 5, e (2009). Beroukhim, R. et al. The landscape of somatic copy-number alteration across human cancers. Nature 463, 899–905 (2010).
Correlation analysis Reveal the relationships between different molecular layers – The strength of association indicates in trans-regulation.
miRNA
GSE10843 GSE10833 microRNA miRNA-mRNA correlation miRNA-ratio correlation miRNA-protein correlation mRNA decay Translational repression Combined effect Association of sequence features with estimated mRNA decay or translation repression Site type Site location Local AU-context Additional 3’ pairing Significant inverse Correlation (p<0.005) Supported by TargetScan, miRanda or MirTarget2 microRNA-target interactions 7235 functional relationships Binding evidence 580 interactions 60miRNAs 423 genes Sequence features on site efficacy microRNA-target interactions mRNA i protein/mRNA ratio protein the relative contribution of translation repression 79 miRNAs 5144 genes Integrative method
Features on site efficacy for these two regulation types mRNA decay : 8mer is efficient Tanslational repression : 8mer site do not show significant efficacy mRNA decay : 3’UTR>ORF>5’UTR translational repression : marginal significance in ORF
Features on site efficacy for these two regulation types AU-rich context appears to favor both mRNA decay and translational repression 3’ pairing enhance mRNA decay, but disfavor efficacy for translational repression
miRNA-target Interactions 60 miRNAs, 423 genes 580 interactions, in which 332 (57.2%) was discovered by the integration of proteomics data miRNA-mRNAmiRNA-ratio miRNA-protein miRNA-mRNA TargetScan miRanda MirTarget2 miRNA-ratio miRNA-protein Function Sequence
miR-138 prefers translational repression SW620 and SW480 (derived from the same patient) SW620SW480 sourcelymph nodeprimary metastasishighpoor miR-138 (log 2 )
Estimate the strength of association between different data Predict the outcome by modeling the combined effect of multiple types of data Linear regression
Ridge—L2 penalized Lasso—L1 penalized Elastic net—L1+L2 penalized
Clustering Unsupervised clustering of omics data to find inherent structures – Using common latent variables among all data types
Network based analysis --using inferred networks or known network interactions to guide analysis
Illustrative example of SNF steps The advantage of the integrative procedure is that weak similarities (low-weight edges) disappear, helping to reduce the noise, and strong similarities (high-weight edges) present in one or more networks are added to the others. Additionally, low-weight edges supported by all networks are retained depending on how tightly connected their neighborhoods are across networks.
Patient similarities for each data types compared to SNF fused similarity
Comparison of SNF with icluster and concatenation
Methods
Extension to more than 2 data types
Tools Sequential or overlap analysis Clustering – R package icluster, iclusterPlus Correlation based Linear regression – – R package glmnet Network based – R package SNFtool Bayesian …..
Visualization: Circular map for omics data Chen et al. Cell 2012, 148(6):
Circos plot Circos Rcircos OmicCircos
IGV
NetGestalt
Things to be aware The importance The challenge in integrative analyses – Dimensionality Integration attempts are best carried out using known biological knowledge
References Kristensen VN. et al. Principles and methods of integrative genomic analyses in cancer. Nat Rev Cancer. 2014, 14(5): Wang B, et al. Similarity network fusion for aggregating data types on a genomic scale. Nat Methods. 2014,11(3): Yuan Y, et al. Assessing the clinical utility of cancer genomic and proteomic data across tumor types. Nat Biotechnol Jul;32(7): Shen R, et al. Integrative clustering of multiple genomic data types using a joint latent variable model with application to breast and lung cancer subtype analysis. Bioinformatics Nov 15;25(22): Liu Q, et al. Integrative omics analysis reveals the importance and scope of translational repression in microRNA-mediated regulation. Mol Cell Proteomics. 2013,12(7): Setty M, et al. Inferring transcriptional and microRNA-mediated regulatory programs in glioblastoma. Mol Syst Biol. 2012;8:605 Lappalainen T. et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature 2013, 501, 506–511 Jacobsen A, et al. Analysis of microRNA-target interactions across diverse cancer types. Nat Struct Mol Biol. 2013, 20(11):