Sorts the summary results of a Messina analysis in decreasing order of classifier margin, and displays the top n. The full sorted data.frame is invisibly returned. If n == 0, displays nothing, but still invisibly returns the full data.frame.
messinaTopResults(result, n = 10)
the result returned by a call to messina, messinaDE, or messinaSurv.
the maximum number of top hits to display (default 10). If zero, no results are displayed, but the full data.frame of results is still returned.
The displayed data.frame has the following columns. Users are encouraged to consult the vignette for a tutorial on how to interpret these results for classification and gene expression tasks.
The feature ID. If the data x supplied to the messina function was an ExpressionSet object, the featureName of the relevant feature of x. Otherwise, if x was a matrix with row names, the row name of the corresponding entry in the matrix, or if x was a matrix without row names, F<n>, where <n> is the row number of the corresponding row of x.
Logical: when its performance was assessed by bootstrapping, did this feature pass the user-supplied performance requirements on out-of-bag data?
A string indicating the type of single-gene classifier that Messina fit to this feature. Valid values are given below, but for most users only the Threshold type is relevant, with the others being only of diagnostic relevance.
A threshold classifer: samples with feature signal at or below the threshold are in one group; samples with feature signal above the threshold are in the other. This is the main result of interest in a Messina analysis, and other classifier types are more of diagnostic interest.
A random (also known as Zero-Rule) classifier. In this case, the feature did not contain sufficient information to construct a good classifier, but the performance requirements were so lenient that simple guessing of an unknown sample's class based on marginal probabilities was enough to satisfy them. The presence of these 'fits' in the top results is indicative of too lenient performance requirements, or a dataset with no predictive value for the classes of interest (at least for single-feature threshold classifiers).
All samples are always called as a single class, and this strategy is sufficient to satisfy the supplied performance requirements. Similar to the "Random" type, the presence of these results are an indicator of too lenient performance requirements.
The feature was not successfully fit. Seen as an indicator of failed fitting in MessinaSurv analyses only, where the Random and OneClass defaults are not applicable.
For a Threshold classifier, the value of the optimal threshold selected by the algorithm. This is the value to use as a cutoff in separating the samples into two classes, either "Group 0" and "Group 1", or "Long surviors" and "Short survivors".
The direction of the threshold classifier. Can take values of either -1 or 1. If -1, samples with expression above the threshold are in group 1 (/TRUE), or have shorter survival times. If 1, samples with expression value above the threshold are in group 0 (/FALSE), and have longer survival times.
The value of the threshold classifier's margin. This is the primary measure of fit strength in a Messina analysis: a higher margin indicates stronger robustness to noise and experimental variations in a classification context, and a higher likelihood of differential expression in a gene expression context.
(invisible) the full table of hits, as a data.frame sorted in order of decreasing margin.
Mark Pinese [email protected]
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
## Load some example data library(antiProfilesData) data(apColonData) x = exprs(apColonData) y = pData(apColonData)$SubType ## Subset the data to only tumour and normal samples sel = y %in% c("normal", "tumor") x = x[,sel] y = y[sel] ## Find differentially-expressed probesets. Allow a sample misattribution rate of ## at most 20%. fit = messina(x, y == "tumor", min_sens = 0.95, min_spec = 0.85) ## Print the 20 probesets with the strongest evidence for differential expression ## between tumour and normal. Save the full table of summary results for later use. summary_table = messinaTopResults(fit, 20) ## Access the top five probesets in the table summary_table[1:5,] ## Examine the summary results for particular probes summary_table[c("204719_at", "207502_at"),]
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.