Shelly’s Notebook: Tues. Oct. 29, Geoduck DMR filtering

Performing group stats on DMRs

Comparing ANOVA vs. GLM significant DMRs

  • ANOVA and GLM each identify different DMRs so I plotted these using an uncorrected p-value < 0.1. Column color bar in heat maps below: cyan = ambient, light pink = low.pH, magenta = super.low.pH. heatmap color: Red = more methylation, blue = no methylation, black = no data.
  • ANOVA significant day 10 MCmax10 DMR heatmap: day10_MCmax10DMR_aov0.1_heatmap.jpg
  • GLM significant day 10 MCmax10 DMR heatmap: day10_MCmax10DMR_glm0.1_heatmap.jpg
  • ANOVA significant day 10 MCmax10 DMR violinplots: day10_MCmax10DMR_aov0.1_boxplots.jpg
  • GLM significant day 10 MCmax10 DMR violinplots: day10_MCmax10DMR_glm0.1_boxplots.jpg
  • CONCLUSIONS:
    • the GLM seems more likely to identify DMRs as significant when one group has zero % methylation and/or half of another group’s samples have zero % methylation.
    • I feel alittle more confident going with the ANOVA

Check ANOVA significant DMRs found by other DMRfind parameters

  • Next I ran ANOVA DMRs from DMR parameters MCmax = 25bp, 30bp, and 50bp for day10 samples to see if these DMRs show obvious group patterns in the heatmaps and violinplots
  • Rmarkdown files that generated figures below are here:
  • ANOVA significant day 10 MCmax25 DMR heatmap: day10_MCmax25DMR_aov0.1_heatmap.jpg
  • ANOVA significant day 10 MCmax25 DMR violinplots: day10_MCmax25DMR_aov0.1_boxplots.jpg
  • ANOVA significant day 10 MCmax30 DMR heatmap: day10_MCmax30DMR_aov0.1_heatmap.jpg
  • ANOVA significant day 10 MCmax30 DMR violinplots: day10_MCmax30DMR_aov0.1_boxplots.jpg
  • ANOVA significant day 10 MCmax50 DMR heatmap: day10_MCmax50DMR_aov0.1_heatmap.jpg
  • ANOVA significant day 10 MCmax50 DMR violinplots: day10_MCmax50DMR_aov0.1_boxplots.jpg
  • CONCLUSIONS: MCmax30 give the most number of DMRs while sill maintaining clear patterns in the heatmap and violinplots. I think the patterns are less obvious in the MCmax50 heatmap. Therefore, I think it’s safe to go with the MCmax30 parameter.
  • NEXT STEPS: Run ANOVA, generate heatmaps and violinplots for DMRs from all 4 comparisons

from shellytrigg https://ift.tt/330ac0J
via IFTTT