From Array Suite Wiki
Differentially Expressed Isoforms
This function gives users the way to identify differentially expressed isoforms between disease and control group. Based on transcript level data, either RPKM, FPKM or Count data, the function converts the expression values to "transcript proportions", dividing the value of each transcript by sum of all transcripts in the same gene. For each transcript, the per-sample proportions will be compared between case and control using a T-test to identify transcripts for which the proportions differ between groups.
By default, it finds all data points (samples) in treatment group having 20% difference from the mean of the control group, then calculates the p-value (using T-test) between the outliers and the control group. It is in the same fashion as the Cancer Outlier Profile Analysis (COPA) in microarray data analysis.
- The window includes a dropdown box to select the Project and Data object on which the command will be run.
- Selections can be made on which variables (transcripts) should be included in the test (options include "all", "selected", "visible", and any pre-generated Lists).
- Selections can also be made on which observations should be included in the Paired Group Test. (options include "all", "selected", "visible", and any pre-generated Lists).
- Gene can be selected from the dropdown box (and comes from columns in the Annotation Table) specify the gene annotation column;
- The Group can be selected from the dropdown box (and comes from columns in the Design Table). Usually, it is a column describing the tumor/normal, or disease/normal.
- The "Compare to" can be selected from the dropdown box (and comes from the Group column in the Design Table). It specifies the normal samples that is to be compared to.
- Minimal difference %: default is 20, for each transcript, finds all data points (samples) in treatment group having 20% difference from the mean of the control group. P value are calculates between these outliers and the control group.
- Minimal expression: the minimal expression cutoff applying to both normal and tumor. The recommendation is 10 reads for count data and 0.1 for RPKM/FPKM values.
Clicking Submit will generate a Table in the Inference tab of the Solution Explorer for the requested test.
The report is on transcript level annotated with gene/transcript annotation. Here are description for key columns:
- RatioDifference: ratio of the transcript in effective tumor samples - Ratio in normal samples
- SampleCount: the total number of tumor samples
- EffectiveSampleCount: number of tumor samples with the expression ratio of transcript > 120% of that in average normal samples
- Pvalue: from t-test of effective tumor samples VS all normal samples.
- GeneExpressionRatio: ratio of expression on gene level between effective tumor samples and all normal samples
- Tumor.Ratio: the average ratio values for transcript in effective tumor samples
- Normal.Ratio: the average ratio values for transcript in normal samples
- Tumor.Expression: the average expression values for transcript in effective tumor samples
- Normal.Expression: the average expression values for transcript in normal samples