RNA-seq Data Analysis Overview
5 likes2,345 views
The document discusses RNA-seq, detailing essential aspects like quality control, alignment, transcript quantification, and differential expression. It emphasizes the importance of experimental design, biases in data, and suggests multiple visualization methods for data analysis. Additional applications and resources are provided, highlighting the need for biological replicates in studies.
RNA-seq Data Analysis Overview
- 1. RNA-seq: Generating Views of the Transcriptome Sean Davis, M.D., Ph.D. Genetics Branch, Center for Cancer Research National Cancer Institute National Institutes of Health
- 2. Normal Karyotype Tumor Karyotype
- 4. Patient and Population phenotype Characteristics Gene Copy Number Sequence Variation Chromatin Structure and Function Gene Expression Transcriptional Regulation DNA Methylation
- 10. Overview • Quality Control • Alignment and Assembly • Transcript Quantification • Visualization • Differential Expression • Experimental Design 10
- 13. Approaches to RNA-seq Nature Biotech (2010) 28, 421-423
- 14. Quality Control • Specialized RNA-seq quality control software • Samples should be “similar” • No “absolute” cutoffs for good vs. bad samples • Visualize data in as many ways as necessary (browser, plots, sample similarity plots like MDS, etc.) 14
- 15. Alignment
- 20. Models for RNA-seq • Count-based models • Multi-reads (isoform resolution) • Paired-end reads (include length resolution step) • Positional bias along transcript length • Sequence bias
- 21. Read Counting
- 22. L. Pachter (2011) arXiv:1104.3889v
- 23. An Example of Sequencing Bias Hansen (2010), NAR
- 34. Distance Metrics Euclidean distance Manhattan distance Minkowski distance (generalized distance)
- 35. Hierarchical Clustering Gene 1 Gene 2 Gene 3 Gene 4 Gene 5 Gene 6 Gene 7 Gene 8
- 36. Hierarchical Clustering Gene 1 Gene 2 Gene 3 Gene 4 Gene 5 Gene 6 Gene 7 Gene 8
- 37. Hierarchical Clustering Gene 1 Gene 2 Gene 3 Gene 4 Gene 5 Gene 6 Gene 7 Gene 8
- 38. Hierarchical Clustering Gene 1 Gene 2 Gene 3 Gene 4 Gene 5 Gene 6 Gene 7 Gene 8
- 40. MA Plot
- 46. RNA-seq workflow as proposed by Anders et al. in Nature Protocols
- 50. Fusion Detection
- 51. Other Applications • Alternative splicing • Isoform utilization • Functional annotation of genomic regions • Allele-specific expression • eQTL analysis • Classification problems (eg., cancer with unknown primary) • … 51
- 52. Experimental Design • What are my goals? – Differential expression? – Transcriptome assembly? – Identify rare, novel trancripts? • System characteristics? – Large, expanded genome? – Intron/exon structures complex? – No reference genome or transcriptome
- 53. Experimental Design • Technical replicates – Probably not needed due to low technical variation • Biological replicates – Not explicitly needed for transcript assembly – Essential for differential expression analysis – Number of replicates often driven by sample availability for human studies – More is almost always better
- 54. Take Home Messages • Defining the experimental question(s) is critical • No gold-standard analysis workflows exist yet • Be aware that experimental biases present in nearly all -omics datasets • Biological replicates are almost always beneficial (necessary) • RPKM/FPKM are for human consumption, not computation (generally) 54
- 55. Links of Interest • http://bioconductor.org • http://biostars.org • http://www.rna-seqblog.com/ • https://genome.ucsc.edu/ENCODE/ • http://www.ncbi.nlm.nih.gov/gds/
Editor's Notes
- #2: I am going to spend a few minutes illustrating how existing and emerging high-throughput genomic technologies are being used to understand cancer, a mindnumbingly complex and disregulated biologic process.
- #3: The first karyotypes were produced in 1956. Shown here is a comparison of a normal karyotype of a normal female and one from a tumor. By 1960, a karyotype of a cancer genome revealed the presence of the Philadelphia chromosome. Now known to represent the BCR-ABL fusion protein, it was not until 33 years later in 1993 that a drug, gleevec, become available that targeted the fusion product. By applying high-throughput microarray technologies, the Cancer Genetics Branch is striving to make observations of the cancer genome that will provide deeper understandings of the biology of cancer, to develop prognostic and diagnostic markers to improve patient-specific treatments, and to find promising targets for directed drug therapy.
- #5: Since Knudson’s famous hypothesis proposing the two-hit model, our understanding of cancer as a genetic disease has progressed to the realization that cancer is not often a function of a single gene gone awry, but probably represents a complex interaction of multiple processes in the genome including altered copy number, gene expression, transcriptional regulation, chromatin modification, sequence variation, and DNA methylation. It is vital to the goal of producing better patient outcomes to understand not only what genes are involved in a certain type of cancer, but also how these other processes affect gene regulation. In short, an integrated view of the cancer genome is necessary and is now becoming possible.