Five Tips for Data Analysis

Five tips for quicker and easier data analysis

We often discuss with clients and users about how to best apply Qlucore Omics Explorer to their specific challenges. We would now like to share with you some of our favorite tips when approaching a new data set.

1.    Use the PCA plot as the first step in analysis to get a quick feeling for if there are any strong signals or patterns in data. Depending on the type of study, patterns can be visible initially, or after some variance filtering. This will quickly give a very useful overview of data. In addition to the PCA plot it is also effective to switch to the table view and look at the real data values to be sure that import and normalization have calculated values in the expected range.

2.    Refine and formalize the exploration mentioned in Tip 1. Use PCA plot, Projection Score and variance filtering to identify structure and patterns. Do this even if the study was started based on a clear hypothesis. Capture the structure using a sample annotation and then visualize the exploratory sample annotation as well as the other annotations describing the underlying hypothesis in a heatmap. The heatmap allows you to visualize as many sample annotation as you like at the same time, and patterns are easily detected. If the exploratory annotation is identical to the experiment design, then data is in good shape and work can continue. If there are differences for some samples then these may be outliers and - should be studied in detail. If there are big discrepancies unexpected signals are present. The reasons could be uncontrolled variables or perhaps new findings.

3.   Use the inbuilt GSEA workbench. When data matches the hypothesis it only requires a few key presses to continue with GSEA analysis and the functional information is presented and available for deeper biological analysis. 

4.    Use the box plot to learn more about individual variables. When a list of interesting (discriminating) variables is identified the flexibility of the box plot saves time. By selecting up to 64 variables at a time the corresponding box plots are generated and give a quick and detailed overview of how the variables are regulated in the subgroups. If the selection of the variables is done in a variable list it is possible to use the key board arrows to scroll down the list and a new box plot is generated. With this method scrolling through hundreds of interesting variables is possible in minutes.

5.   Many users miss the Fold change slider. It is found at the lower part of the statics dialog and makes it easy to combine fold change with statistical filtering.

Finally, the suite of video tutorials are between 4 to 6 minutes long and go through a lot of the above mentioned functionality. Use them, they will save you time.

Good luck with the exploration and your analysis!