Presentation is loading. Please wait.

Presentation is loading. Please wait.

Assessing changes in data – Part 2, Differential Expression with DESeq2

Similar presentations


Presentation on theme: "Assessing changes in data – Part 2, Differential Expression with DESeq2 "β€” Presentation transcript:

1 Assessing changes in data – Part 2, Differential Expression with DESeq2
𝑇=1 𝑇=2

2 DESeq2 model and normalization
A negative binomial GLM is then used to model the count data with a log link applied directly to the true concentration: 𝑦 𝑔𝑖 ~ NB( mean=πœ‡ 𝑔𝑖 , dispersion= 𝛼 𝑔 ) log π‘ž 𝑔𝑖 = π‘Ÿ π‘₯ π‘–π‘Ÿ 𝛽 π‘”π‘Ÿ mean: πœ‡ 𝑔𝑖 = 𝑠 𝑖 π‘ž 𝑔𝑖 where π‘₯ π‘–π‘Ÿ is the design matrix and 𝛽 π‘”π‘Ÿ relates to the log fold change of gene 𝑔 Where π‘ž 𝑔𝑖 represents the true concentration of fragments from gene 𝑔 in sample 𝑖, and 𝑠 𝑖 is the size factor (normalization), given by: e.g., for 2-level (contrl vs. treatment) experiment, with 2 replicates each: π‘₯ π‘–π‘Ÿ = 𝑠 𝑖 = median 𝑔 𝑦 𝑔𝑖 𝑦 𝑔 𝑅 𝛽 𝑔 = 𝛽 𝑔0 𝛽 𝑔1 𝑦 𝑔 𝑅 = 𝑖=1 π‘š 𝑦 𝑔𝑖 1/m with: Log fold-change result from GLM fit: LFC g = (𝛽 𝑔1 βˆ’ 𝛽 𝑔0 ) (geometric mean of counts for gene 𝑔)

3 Our data for today We have nascent transcription sequence data from 4 libraries generated from human cells??: control – 2 biological replicates treatment – 2 biological replicates Control condition: -/- p53 knockout treated with DMSO (measured at 30 minutes)?? Treatment condition: -/- p53 knockout treated with Nutlin (measured at 30 minutes)?? Our features: A continuous region for each gene that including TSS, exons, and introns. Feature

4 Steps for DE analysis using DESeq2
Load in/generate matrix of count data Define your sample table with treatment conditions Determine the structure of your design matrix Build your DESeq2 data structure (DESeqDataSet) Filter out very lowly expressed features (optional) Run the DESeq() method (estimates dispersion, fits GLM, calculates LFC) Define your alpha value Extract significant results Perform LFC shrinkage (for purposes of visualization) Plot results Now lets go through each of these steps in our script, one at a time…

5 0) First, open a terminal, cd to the location of your count data, and run R

6 Load data Define sample conditions

7 3) Determine your design matrix
4) Build your DESeqDataSet data structure

8 5) Filter out lowly expressed features
6) Run DESeq()

9 At this point, let’s look at the dispersion estimates

10 7) Define your alpha value
8) Extract significant results 9) Perform LFC shrinkage

11

12 10) Plot your results

13 10) Plot your results

14 Lastly, sort results by significance and filter

15 DESeq2 for transcript (gene isoform) abundance
tximport package from bioconductor SalmonΒ (Patro et al. 2017) SailfishΒ (Patro, Mount, and Kingsford 2014) kallistoΒ (Bray et al. 2016) RSEMΒ (Li and Dewey 2011)

16


Download ppt "Assessing changes in data – Part 2, Differential Expression with DESeq2 "

Similar presentations


Ads by Google