Dahlquist:Notebook/Microarray Data Analysis/2008/10/21

From OpenWetWare

Jump to: navigation, search

Image:HeaderBlueBig.gif Home        Research        Protocols        Notebook        People        Publications        Courses        Contact       


Microarray Data Analysis Main project page
Previous entry      Next entry

Today's Workflow

The results generated on 10/14/2008 were downloaded and placed on the Desktop in "Edge Analysis" in Kevin's profile. Significant gene results were saved as tab-delimited files and the Pvalue Histograms and QPlots were saved into a powerpoint and printed.

  • Only the wt-only results should be used, the other results are useless, see below for explanation.

Previous run (10/14/2008) on dCIN5-only dataset gave interesting results. While the wt-only dataset produced about 1000 significant genes, the dCIN5-only one gave about 150 significant genes. To verify this result:

  • First the covariates and genelist files were uploaded to lion share. They will be opened with excel and checked for errors.
    • IMPORTANT: It was found that the flask numbers were wrong for covariates files for dCIN5-only and wt-vs-dCIN5. They were changed and new runs were completed.
    • The new files were saved on the desktop in the Edge Analysis folder as:
      • dCIN5-only_Edge_covariates_20081021.txt and
      • wt-vs-dCIN5_Edge_covariates_20081021.txt

Reran the dCIN5-vs-wt data with the updated covariate file:

  • Gene file in Desktop "Data analysis 2008-10-02", Covariate file on Desktop
    • Used gene file "wt-dCIN5_consolidated_Edge_genes-indexonly_20080715.txt"
    • Used covariate file "wt-dCIN5_consolidated_Edge_covariates_20081021.txt
  • Load both into an Edge session.
  • Select "Impute Missing Data" from the menu. Calculate Percent Missing Data by clicking on the button. The results are:
    • Percent of genes missing data: 7.63%
    • Percent of arrays missing data: 95.35%
    • Overall percent of missing data: 3.15%
  • For KNN Parameters, set:
    • Percent of missing values to tolerate in a gene: 100 (so all genes included)
    • Number of nearest neighbors to use (maximum of 15): 15
    • clicked GO to impute missing data.
  • Selected "Identify Differentially Expressed Genes"
    • Class Variable is: Strain
    • Differential Expression Type is: Time Course
    • Number of null iterations, set to 1000
    • Choose a seed for reproducible results, set to 47
    • Choose Time Course Settings
    • Covariate giving time points is: Timepoint
    • Covariate corresponding to individuals is: Flask
    • Choose spline type, accepted default of Natural Cubic Spline, dimension 4
    • Click "Apply" and then click "Go"
    • 1000 permutations looks like it will take about 9 minutes.
  • Results: (Saved in 2008-10-14 Results)
    • 2 significant genes under these settings.(ID 1068 and 1798) with Q Value Cutoff of 0.1
    • Choose show all
      • Saved total list of genes as: "GeneList_20081021_wt-vs-dCIN5"
    • To save the plots, do the following command in the R console window.
savePlot(filename = "PvalHistogram_wt-vs-dCIN5", type = c("png"), device = dev.cur())
  • This will save the active plot window under a file name you choose. Saves in folder "edge_1.1.290"
    • Saved Q-Plot as "QPlot_20081021_wt-vs-dCIN5"
    • Saved Histograms as "PvalHistogram_20081021_wt-vs-dCIN5

Then dCIN5 dataset was ran on its own:

  • Gene file in "Edge_data_20080710" and covariate file on Desktop
    • Used gene file "dCIN5-only_Edge_genes-indexonly_20080715.txt"
    • Used covariate file "dCIN5-only_Edge_covariates_20081021.txt"
  • Load both into an Edge session.
  • Select "Impute Missing Data" from the menu. Calculate Percent Missing Data by clicking on the button. The results are:
    • Percent of genes missing data: 1.32%
    • Percent of arrays missing data: 90%
    • Overall percent of missing data: 0.09%
  • For KNN Parameters, set:
    • Percent of missing values to tolerate in a gene: 100 (so all genes included)
    • Number of nearest neighbors to use (maximum of 15): 15
    • clicked GO to impute missing data.
  • Selected "Identify Differentially Expressed Genes"
    • Class Variable is: None (within class differential expression)
    • Differential Expression Type is: Time Course
    • Number of null iterations, set to 1000
    • Choose a seed for reproducible results, set to 47
    • Choose Time Course Settings
    • Covariate giving time points is: Timepoint
    • Covariate corresponding to individuals is: Flask
    • Choose spline type, accepted default of Natural Cubic Spline, dimension 4
    • Click "Apply" and then click "Go"
    • 1000 permutations looks like it will take about 2 minutes.
  • Results: (Saved on Desktop)
    • 1000 Genes Called Significant (Cutoff Q Value 0.0114)!!!!
    • Saved total list of genes as "GeneList_20081021_dCIN5-only"
    • Saved Q-Plot as "QPlot_20081021_dCIN5-only"
    • Saved Histograms as "PvalHistogram_20081021_dCIN5-only"

All Q-Plots and Pvalue Histograms were combined into a powerpoint. All significant gene lists were exported and saved into text files in Edge Analysis on the Desktop in Dahlquist's Lab.



Personal tools