Help : Data Plotter Help
Data Plotter:
This tool can be used to visualize your data and to decide how to filter it during analysis. It can be used for an individual experiment or for a batch of experiments.To access the Data Plotter for a single experiment, navigate to Advanced Search, display a list of experiments and then click on the Data Plotter icon:
adjacent to the experiment for which you want to plot data.
If you want to compare plots for all or some experiments in an experiment set, navigate to Basic Search, select the experiment set name, click on "Display Data" and then "Display Data" again, and then click on the data plotter icon next to the set name on the following page. You will see a page on which you can de-select some of the experiments if you don't want every one plotted, can select the graph type and can filter the data. After clicking "Display", a page with the plots for each experiment will be produced.
You can choose to plot the data as a Histogram or as a Scatter Plot.
Histogram
(a bar graph that shows frequency data):
This graph will show the distribution of values for one or two columns simultaneously.For example, to visualize the distribution(frequency vs. value) of the Log(base 2) of R/G Normalized Ratio(Mean) values for an experiment:
For Histogram plots, the values will be plotted along the x-axis regardless of whether you select the x and/or y axis. The frequency will be plotted on the y-axis. Pick either axis, select "Field" = Log(base2) of R/G Normalized Ratio(Mean) and leave the "Active Filter" checkbox checked. If you select to plot on a linear scale, base will be irrelevant."Choose the type of plot to make:" = Histogram
"Axis Plot?" :
Uncheck the other box unless you want to see another value plotted on the same graph.Then decide which spots to include in the histogram: whether you want to include control spots, empty spots and/or flagged spots and whether you want to exclude any other data by making filters active(by selecting a measurement and value and then checking the "Active Filter" checkbox). All "Active Filters" will be combined with a logical AND unless you use the logic text box to specify otherwise.
Click on "Display" and you will get a frequency vs. value bar graph.
One interesting use for this tool is to view normalized vs. unnormalized data for an experiment.
Normalization
You also may want to see the shape of the distribution of the data.
You can plot your data in scatter plot format to visualize how the values are correlated.Visualizing your data in scatter plot format can also be useful for selecting filters to apply before you cluster your data. For example, if you want to select a good cutoff value to filter data based on regression correlation (which uses the pixel to pixel correlation within a spot as a measure of within-spot consistency and corresponding data reliability), you can plot the Log(base 2) of R/G Normalized Ratio(Mean) vs. Regression Correlation. Look for a point in the regression correlation values where above that value, the data appear well-behaved, and where below that value you start to see problems.
In general, the log ratios should center around 0 and not vary with decreasing regression correlation values. In this example, the log ratios appear to diverge below a regression correlation value of about 0.4 - 0.6. If you set a regression ratio filter cutoff at 0.6 for data retrieval, you will leave out these probably unreliable spots.
Another interesting plot to view is Log(base 2) of R/G Normalized Ratio(Mean) vs. Log 10(Channel 2 Normalized(Mean Intensity/Median Background Intensity)). Try this for both the red and green channels to decide where to set filters to eliminate spots having a low signal/noise ratio.
At low relative intensity, the log ratios don't center around zero. Use a cutoff value at about 0.3 on the x-axis (the Log(base 10) Ch2 foreground/background ratios) to filter these spots from your analysis.
If you plot the Log(base 2) of R/G Normalized Ratio(Mean) vs. Log(2)(Ch2 Normalized Net(Mean)), you typically get a fish shaped distribution.
Generally the "tail" is technical noise and the true data values are zero. You can select a filter value to leave the spots comprising the "tail" out of your analysis. If you plot the Log(2)(Ch2 Normalized Net (Mean)) vs. Log(2)(Ch1 Net (Mean)), it can be used as a quality control measure in that the graph can show a sharp cutoff point on the right.
This indicates a saturation point for fluorescence which is below the technical saturation level (where the spots would show up as white in your scanned images) but at which sensitivity has been lost in your data. In this type of plot, you generally want to see the data evenly spread along the x and y axes. A separation into two populations may indicate technical problems with the microarrays.
Please send comments or questions to: array@genome.stanford.edu
