bklein7
commented
6 years ago The weighted networks for db2-db6 were also color coded to reflect average log2 expression changes at 15 minutes and 60 minutes. A new file containing the results for db1-db6 was uploaded (https://github.com/kdahlquist/DahlquistLab/blob/master/data/Fall2017/db1-db6_log2-expression-coding_BK20171205.pptx).
Color coding nodes by frequency of significant ANOVA p-values and log2 expression changes at 15 and 60 minutes aided in the interpretation of weighted networks. ANOVA color coding visually demonstrated that the majority of nodes in all networks, with the exception of the smaller ∆cin5-derived network (6/14 genes had significant ANOVA p-values), exhibited significant expression changes during cold shock. This supports our modeling efforts. Further, coding by expression change facilitated interpretation of network dynamics and was generally consistent with expected regulation based on weighted inputs. Although some nodes exhibited expression changes that did not appear consistent with their inputs (e.g. SWI4 and SWI5 in db4), these nodes tended not to have significant ANOVA p-values. Thus, if a similar color coding feature based on expression is implemented in GRNsight, it may be desirable not to color nodes with insignificant (B&H) ANOVA values, as was the standard used by GenMAPP.
The need for considering ANOVA p-values and changes in expression of individual nodes within db1-db6 was discussed, as model interpretations are not meaningful when modeling noise. To tackle this issue, I compiled ANOVA B&H p-values for all genes and strains represented in db1-db6 (https://github.com/kdahlquist/DahlquistLab/blob/master/data/Fall2017/ANOVA-B%26H-p-values_all-genes-and-strains_BK20171205.xlsx). Nodes with significant p-values were separated into categories and db1-db6 were color coded based on these findings (https://github.com/kdahlquist/DahlquistLab/blob/master/data/Fall2017/db1-db6_ANOVA-color-coding_BK20171205.pptx).
A similar protocol was followed to visualize the expression change of individual nodes in db1-db6. It had been previously mentioned that formatting the nodes in weighted networks as in GenMAPP could be useful. Average log2 expression changes at 15 minutes and 60 minutes (average of all time points for all strains) were calculated for the nodes in all networks and color coded (https://github.com/kdahlquist/DahlquistLab/blob/master/data/Fall2017/log2-expression-coding_db1-db6_BK20171205.xlsx). The weighted network for db1 was used as a sample, and individual nodes were color coded to convey expression changes in the early and late responses to cold shock (https://github.com/kdahlquist/DahlquistLab/blob/master/data/Fall2017/log2-expression-coding_db1_BK20171205.pptx). The weighted network for db1 generated based on significant ANOVA p-values was also included for reference. Moving forward, this format could be modified and extended to db2-db6.