Version 1.1.7

• Fixed an issue where statistical analysis results would not display properly for files with only one comparison after a missing value filter was applied.

Version 1.1.6

• Fixed an issue where, with certain input files, the raw average expression/raw fold change columns were not sorted properly in the output data files.

Version 1.1.5

• Fixed an issue where columns were erroneously named if one column name was a substring of another column name. This fix required re-naming some of the raw expression value columns.
• Replaced the zero-centered model coefficients in the moderated F test table with the normalized average expression values.
• Removed pairwise information from moderated F test table.
• Generalized profile plot axis labels.

Version 1.1.4

• Fixed an issue with the one-sample and two-sample T-test. Previously, in the rare case where a protein was not detected in all samples in a certain group, any test comparisons involving that group would be erroneously discarded, which would cause an error when attempting to export the signed log-transformed p-values. Now, all test results are reported for all proteins: NAs are reported when a protein was excluded from a particular test.
• Fixed an issue where exporting the RMarkdown report would fail when attempting to export the heatmap.
• Fixed an issue where a GCT file with only one cdesc column could not be imported correctly.
• Fixed some column descriptors in output GCT files.
• Updated various help text within the app.

Version 1.1.3

• Occasionally, eBayes(trend=TRUE) fails for intensity-based data, particularly when the distribution of quantified features is not uniform across samples. In these cases, eBayes(trend=FALSE) is run instead, and a warning message is printed. We highly encourage users who encounter this warning to carefully examine their data, and re-perform statistical analysis as needed. Typically, setting a stricter missing value filter will fix the issue.
• The one-sample T-test is now fixed. It was not working due to an error in passing a parameter to the function.
• An issue with heatmap color annotations was fixed. The issue occured during conversion of certain annotation group names to R-compliant names.

Version 1.1.2

• Previously, raw log-fold changes were median-centered by default. Raw log-fold changes are no longer median-centered so that they accurately reflect the fold change before any normalization is performed.
• When a one-sample T-test or no test is performed, raw average abundances per group are reported in the place of fold change. Column headers and descriptors in the exported excel file have been updated to reflect this for these cases.
• Help text has been updated to clarify that data must be log-transformed prior to statistical analysis. The statistical testing performed in Protigy is not valid for non-log-transformed data.

Version 1.1.1

• A checkbox has been added for group-wise normalization on the group assignment screen. When this box is checked, group normalization will be performed based on the selected column. When the box is left unchecked, group normalization will not be performed. This replaces the previous checkbox on the statistical analysis screen.
• There is now a help button that explains how the missing value filter works, as well as help text within the application.
• The correlation boxplot is now always scaled from 0 to 1 for positive values. For negative values, the boxplot will scale based on the minimum value. This should produce more comparable boxplots across datasets.
• More information has been added to the workflow summary including if intensity data were used, if group-wise normalization was performed, and if QC.fail samples were filtered.

Version 1.1

• A checkbox has been added to filter out samples where QC.status=QC.fail. This option is for GCT files ONLY. If the QC.status column does not exist and the box is checked, a warning message is printed and analysis will proceed using all samples.
• A warning message appears when attempting to log-transform a dataset that contains negative values (i.e., a dataset that has already been log-transformed). In this case, log-transformation will not occur to prevent downstream analysis errors.
• An error message appears when a column label has only one sample assigned to it, and analysis will not proceed. ProTIGY requires all column labels have more than one sample assigned to them. When this error appears, the user must either remove that sample from the file, or use another column for statistical testing and/or group-wise normalization.
• A checkbox has been added to indicate the data are intensity data rather than ratios. When the box is checked, normalization methods, filtering methods, and statistical testing will be automatically filtered to only methods appropriate for intensity data. Further, the missing value rate will be capped at 99%, as the statistical analysis can fail when the missing value rate is 100%.
• For the moderated two-sample t-test and the moderated F test, eBayes(trend=true) is used for intensity data. This option is more stable for intensity data, but does require the capped missing value rate of 99%. eBayes(trend=FALSE) continues to be used for ratios.
• Help text has been edited for non-GCT files to clarify that NA must be used for missing sample annotations. Further, a sample can be excluded in a non-GCT file by leaving the Experiment and Group annotations as NA.

Version 1.0.4

• Fixes multiple plotting issues.
• Volcano plot labels may now be chosen from the following: ID_Symbol (default), ID, or Symbol.
• Gene symbol support for Ensembl protein IDs is now included.
• For gct files, if a geneSymbol column is included in the row descriptors (rdesc), it is used to determine the symbol rather than using the available database.

Version 1.0.3

Fixes issues with heatmap visualization and export.