| Title: | Classification by local similarity threshold |
|---|---|
| Description: | Package for modified nearest-neighbor classification based on calculation of a similarity threshold distinguishing within-group from between-group comparisons. |
| Authors: | Noah Hoffman |
| Maintainer: | Noah Hoffman <[email protected]> |
| License: | GPL-3 |
| Version: | 1.55.0 |
| Built: | 2024-10-30 04:41:15 UTC |
| Source: | https://github.com/bioc/clst |
Package for modified nearest-neighbor classification based on calculation of a similarity threshold distinguishing within-group from between-group comparisons.
| Package: | clst |
| Type: | Package |
| License: | GPL-3 |
| Author: | Noah Hoffman <[email protected]> |
Index:
Further information is available in the following vignettes:
clstDemo |
clst (source, pdf) |
TODO: write package overview.
Noah Hoffman
Maintainer: <[email protected]>
library(clst) packageDescription("clst") data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species i <- 1 cc <- classify(dmat, groups, dvect=dmat[i,]) cat('query at i =',i,'is species',paste('I.', groups[i]),'\n') printClst(cc) i <- 125 cc <- classify(dmat, groups, dvect=dmat[i,]) cat('query at i =',i,'is species',paste('I.', groups[i]),'\n') printClst(cc)library(clst) packageDescription("clst") data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species i <- 1 cc <- classify(dmat, groups, dvect=dmat[i,]) cat('query at i =',i,'is species',paste('I.', groups[i]),'\n') printClst(cc) i <- 125 cc <- classify(dmat, groups, dvect=dmat[i,]) cat('query at i =',i,'is species',paste('I.', groups[i]),'\n') printClst(cc)
Square matrices decsribing pairwise distances among 16s rRNA sequences.
data(actino)data(actino)
List of 5 $ dmat1 : num [1:146, 1:146] 0 0.763 1.25 10.345 12.771 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:146] "200" "201" "202" "203" ... .. ..$ : chr [1:146] "200" "201" "202" "203" ... $ dmat2 : num [1:146, 1:146] 0 0.574 1.044 5.669 8.409 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:146] "200" "201" "202" "203" ... .. ..$ : chr [1:146] "200" "201" "202" "203" ... $ dmat3 : num [1:146, 1:146] 0 0.763 1.25 8.571 11.233 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:146] "200" "201" "202" "203" ... .. ..$ : chr [1:146] "200" "201" "202" "203" ... $ taxa : Factor w/ 33 levels "Actinomyces bowdenii",..: 12 12 12 23 20 20 8 22 12 20 ... $ abbrev: Factor w/ 33 levels "A bowdenii","A canis",..: 12 12 12 23 20 20 8 22 12 20 ...
The matrices $dmat1, dmat2, and dmat3 contain
percent nucleotide difference with indels penalized heavily, little, and
somewhat, respectively.
$taxa is a factor of species names; abbreviations of the same
names are found in $abbrev.
data(actino)data(actino)
Tree-derived pairwise distances and taxonomic assignments among 16S rRNA sequences representing bacteria represented in the vaginal mucosa.
data(bvseqs)data(bvseqs)
The format is: List of 3 $ dmat : num [1:448, 1:448] 0 0.0494 0.0968 0.1002 0.1606 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:448] "S001098970" "S000859776" "S000539896" "S001352901" ... .. ..$ : chr [1:448] "S001098970" "S000859776" "S000539896" "S001352901" ... $ groupTab:'data.frame': 448 obs. of 12 variables: ..$ superkingdom : chr [1:448] "2" "2" "2" "2" ... ..$ superphylum : chr [1:448] NA NA NA NA ... ..$ phylum : chr [1:448] "1224" "1224" "1224" "1224" ... ..$ class : chr [1:448] "1236" "1236" "1236" "1236" ... ..$ subclass : chr [1:448] NA NA NA NA ... ..$ order : chr [1:448] "72274" "72274" "72274" "72274" ... ..$ suborder : chr [1:448] NA NA NA NA ... ..$ family : chr [1:448] "468" "468" "468" "468" ... ..$ genus : chr [1:448] "469" "469" "469" "469" ... ..$ species_group : chr [1:448] NA NA NA NA ... ..$ species_subgroup: chr [1:448] NA NA NA NA ... ..$ species : chr [1:448] "470" "470" "471" "470" ... $ taxNames: Named chr [1:212] "Actinomyces urogenitalis" "Lactobacillus jensenii" "Proteobacteria" "Gammaproteobacteria" ... ..- attr(*, "names")= chr [1:212] "103621" "109790" "1224" "1236" ...
(Describe creation of this data set)
Sequences were assembled from both the RDP 16S rRNA database and from the laboratory of Dr. David Fredricks.
RDP url here.
data(bvseqs) ## maybe str(bvseqs) ; plot(bvseqs) ...data(bvseqs) ## maybe str(bvseqs) ; plot(bvseqs) ...
Functions to perform classification by local similarity threshold.
classify(dmat, groups, dvect, method = "mutinfo", minScore = 0.45, doffset = 0.5, dStart = NA, maxDepth = 10, minGroupSize = 2, objNames = names(dvect), keep.data = TRUE, ..., verbose = FALSE) classifyIter(dmat, groupTab, dvect, dStart = NA, multiple = FALSE, keep.data = TRUE, ..., verbose = FALSE) classifier(dmat, groups, dvect, method = 'mutinfo', minScore = 0.45, doffset = 0.5, dStart = NA, minGroupSize = 2, objNames = names(dvect), keep.data = TRUE, ..., verbose = FALSE, depth = 1) pull(dmat, groups, index) pullTab(dmat, groupTab, index)classify(dmat, groups, dvect, method = "mutinfo", minScore = 0.45, doffset = 0.5, dStart = NA, maxDepth = 10, minGroupSize = 2, objNames = names(dvect), keep.data = TRUE, ..., verbose = FALSE) classifyIter(dmat, groupTab, dvect, dStart = NA, multiple = FALSE, keep.data = TRUE, ..., verbose = FALSE) classifier(dmat, groups, dvect, method = 'mutinfo', minScore = 0.45, doffset = 0.5, dStart = NA, minGroupSize = 2, objNames = names(dvect), keep.data = TRUE, ..., verbose = FALSE, depth = 1) pull(dmat, groups, index) pullTab(dmat, groupTab, index)
dmat |
Square matrix of pairwise distances. |
groups |
Object coercible to a factor identifying group
membership of objects corresponding to either edge of |
groupTab |
a data.frame representing a taxonomy, with columns in increasing order of specificity from left to right (ie, Kingdom –> Species). Column names are used to name taxonomic ranks. Rows correspond to margins of dmat. |
dvect |
numeric vector of distance from query sequence to each reference corresponding to margins of dmat. |
method |
The method for calculating the threshold; only 'mutinfo' is currently implemented. |
minScore |
Threshold value for the match score to define a match. |
doffset |
Offset used in the denominator of the expression to calculate match score to penalize very small groups of reference objects. |
dStart |
start with this value of |
multiple |
if TRUE, stops at the rank that yields at least one match; if FALSE, continues to perform classification until exactly one match is identified. |
maxDepth |
Maximum number of iterations that will be attempted to perform classification. |
minGroupSize |
The minimal number of members comprising at least one group required to attempt classification. |
objNames |
Optional character identifiers for objects
corresponding to margin of |
keep.data |
Populates |
verbose |
Terminal output is produced if TRUE. |
index |
an integer specifying an element in |
... |
see Details |
depth |
specifies iteration number (not meant to be user-defined) |
classify performs iterative classification. See the vignette
vignette for package clst for a description of the classification
algorithm.
classifier performs non-iterative classification, and is
typically not called directly by the user.
The functions pull and pullTab are used to remove a single
element of dmat for the purpose of performing classification
agains the remaining elements. The value of these two functions (a list)
can be passed directly to classify or classifyIter
directly (see examples).
classify and classifyIter return x, a list
of lists, one for each iteration of the classifier. Each sub-list
contains the following named elements:
depth |
An integer indicating the number of the iteration (where x[[i]]$depth == i) |
tally |
a |
details |
a list of two matrices, named "below" and "above",
itemizing each object with index i in the reference set with
distances below or above the distance threshold D,
respectively. Columns include |
matches |
Character vector naming groups to which query object belongs. |
thresh |
object returned by |
params |
a list of input arguments and their values |
input |
list containing copies of |
Noah Hoffman
## illustrate classification using the Iris data set data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species ## remove one element from the data set and perform classification using ## the remaining elements as the reference set ind <- 1 cat(paste('class of "unknown" sample is Iris',groups[ind]),fill=TRUE) cc <- classify(dmat[-ind,-ind], groups[-ind], dvect=dmat[ind, -ind]) printClst(cc) ## this operation can be performed conveinetly using the `pull` function ind <- 51 cat(paste('class of "unknown" sample is Iris',groups[ind]),fill=TRUE) cc <- do.call(classify, pull(dmat, groups, ind)) printClst(cc) str(cc)## illustrate classification using the Iris data set data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species ## remove one element from the data set and perform classification using ## the remaining elements as the reference set ind <- 1 cat(paste('class of "unknown" sample is Iris',groups[ind]),fill=TRUE) cc <- classify(dmat[-ind,-ind], groups[-ind], dvect=dmat[ind, -ind]) printClst(cc) ## this operation can be performed conveinetly using the `pull` function ind <- 51 cat(paste('class of "unknown" sample is Iris',groups[ind]),fill=TRUE) cc <- do.call(classify, pull(dmat, groups, ind)) printClst(cc) str(cc)
Identify a distance threshold predicting whether a pairwise distance represents a comparison between objects in the same class (within-group comparison) or different classes (between-group comparison) given a matrix providing distances between objects and the group membership of each object.
findThreshold(dmat, groups, distances, method = "mutinfo", prob = 0.5, na.rm = FALSE, keep.dists = TRUE, roundCuts = 2, minCuts = 20, maxCuts = 300, targetCuts = 100, verbose = FALSE, depth = 1, ...) partition(dmat, groups, include, verbose = FALSE)findThreshold(dmat, groups, distances, method = "mutinfo", prob = 0.5, na.rm = FALSE, keep.dists = TRUE, roundCuts = 2, minCuts = 20, maxCuts = 300, targetCuts = 100, verbose = FALSE, depth = 1, ...) partition(dmat, groups, include, verbose = FALSE)
dmat |
Square matrix of pairwise distances. |
groups |
Object coercible to a factor identifying group
membership of objects corresponding to either edge of |
include |
vector (numeric or boolean) indicating which elements to retain in the output; comparisons including an excluded element will have a value of NA |
distances |
Optional output of |
method |
The method for calculating the threshold; only 'mutinfo' is currently implemented. |
prob |
Sets the upper and lower bounds of D as some quantile of the within class distances and between-class differences, respectively. |
na.rm |
If TRUE, excludes |
keep.dists |
If TRUE, the output will contain the |
roundCuts |
Number of digits to round cutoff values (see Details) |
minCuts |
Minimal length of vector of cutoffs (see Details). |
maxCuts |
Maximal length of vector of cutoffs (see Details) |
targetCuts |
Length of vector of cutoffs if conditions met by
|
verbose |
Terminal output is produced if TRUE. |
depth |
Private argument used to track level of recursion. |
... |
Extra arguments are ignored. |
findThreshold is used internally in classify, but
may also be used to calculate a starting value of $D$.
partition is used to transform a square (or lower triangular)
distance matrix into a data.frame containing a column of
distances ($vals) along with a factor ($comparison)
defining each distance as a within- or between-group
comparison. Columns $row and $col provide indices of
corresponding rows and columns of dmat.
In the case of findThreshold, output is a list with elements
decsribed below. In the case of partition, output is the data.frame
returned as the element named $distances in the output of
findThreshold.
D |
The distance threshold (distance cutoff corresponding to the PMMI). |
pmmi |
Value of the point of maximal mutual information (PMMI) |
interval |
A vector of length 2 indicating the upper and lower bounds over which values for the threshold are evaluated. |
breaks |
A |
distances |
If |
method |
Character corresponding to input argument |
params |
Additional input parameters. |
Noah Hoffman
data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species thresh <- findThreshold(dmat, groups, type="mutinfo") str(thresh)data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species thresh <- findThreshold(dmat, groups, type="mutinfo") str(thresh)
link{findThreshold}
The functions plotDistances and plotMutinfo are used to
visualize the distance threshold calculated by
findThreshold in the context of pairwise distances among
objects in the reference set.
plotDistances(distances, D = NA, interval = NA, ylab = "distances", ...) plotMutinfo(breaks, D = NA, interval = NA, xlab = "distance", ylab = "mutual information", ...)plotDistances(distances, D = NA, interval = NA, ylab = "distances", ...) plotMutinfo(breaks, D = NA, interval = NA, xlab = "distance", ylab = "mutual information", ...)
distances |
The |
breaks |
The |
D |
The distance threshold |
interval |
The range of values over which candidiate values of PMMI are evaluated. |
xlab |
Label the x axis of the plot. |
ylab |
Label the y axis of the plot. |
... |
Additional arguments are passed to |
(plotMutinfo)
plotDistances produces a box-and-whisker plot contrasting within-
and between-group distances. plotMutinfo produces a plot of
cutpoints vs mutual information scores.
Returns a lattice grid object.
Noah Hoffman
data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species thresh <- findThreshold(dmat, groups) do.call(plotDistances, thresh) do.call(plotMutinfo, thresh)data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species thresh <- findThreshold(dmat, groups) do.call(plotDistances, thresh) do.call(plotMutinfo, thresh)
Prints a description of the output of classify.
printClst(cc, rows = 8, nameWidth = 30, groupNames)printClst(cc, rows = 8, nameWidth = 30, groupNames)
cc |
Output of |
rows |
Number of rows corresponding to groups of reference objects to show. |
nameWidth |
Character width of group names. |
groupNames |
a named vector containing replacement names for
groups keyed by categories in |
Output value is NULL; output is to stdout.
Noah Hoffman
data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Speciesdata(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species
Produces annotated representations of two-dimensional
multidimensional scaling plots using cmdscale.
scaleDistPlot(dmat, groups, fill, X, O, indices = "no", include, display, labels, shuffleGlyphs = NA, key = "top", keyCols = 4, glyphs, xflip = FALSE, yflip = FALSE, ...)scaleDistPlot(dmat, groups, fill, X, O, indices = "no", include, display, labels, shuffleGlyphs = NA, key = "top", keyCols = 4, glyphs, xflip = FALSE, yflip = FALSE, ...)
dmat |
Square matrix of pairwise distances. |
groups |
Object coercible to a factor identifying group
membership of objects corresponding to either edge of |
fill |
vector (logical or indices) of points to fill |
X |
vector of points to mark with an X |
O |
vector of points to mark with a circle |
indices |
label points with indices (all points if 'yes', or a subset indicated by a vector) |
include |
boolean or numeric vector of elements to include in call to cmdscale |
display |
boolean or numeric vector of elements to include in call to display |
labels |
list or data frame with parameters $i indicating indices and $text containing labels. |
shuffleGlyphs |
modify permutation of shapes and colors given an integer to serve as a random seed. |
key |
'right' (single column), 'top' (variable number of columns), or NULL for no key |
keyCols |
number of columns in key |
glyphs |
a data.frame with columns named |
xflip |
if TRUE, flip orientation of x-axis |
yflip |
if TRUE, flip orientation of y-axis |
... |
additional arguments are passed to |
Returns a lattice grid object.
Noah Hoffman
data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species ## visualize pairwise euclidean dstances among items in the Iris data set fig <- scaleDistPlot(dmat, groups) plot(fig) ## leave-one-out analysis of the classifier loo <- lapply(seq_along(groups), function(i){ do.call(classify, pull(dmat, groups, i)) }) matches <- lapply(loo, function(x) rev(x)[[1]]$matches) result <- sapply(matches, paste, collapse='-') confusion <- sapply(matches, length) > 1 no_match <- sapply(matches, length) < 1 plot(scaleDistPlot(dmat, groups, fill=confusion, O=confusion, X=no_match))data(iris) dmat <- as.matrix(dist(iris[,1:4], method="euclidean")) groups <- iris$Species ## visualize pairwise euclidean dstances among items in the Iris data set fig <- scaleDistPlot(dmat, groups) plot(fig) ## leave-one-out analysis of the classifier loo <- lapply(seq_along(groups), function(i){ do.call(classify, pull(dmat, groups, i)) }) matches <- lapply(loo, function(x) rev(x)[[1]]$matches) result <- sapply(matches, paste, collapse='-') confusion <- sapply(matches, length) > 1 no_match <- sapply(matches, length) < 1 plot(scaleDistPlot(dmat, groups, fill=confusion, O=confusion, X=no_match))
Square matrices decsribing pairwise distances among 16s rRNA sequences.
data(strep)data(strep)
List of 5 $ dmat1 : num [1:150, 1:150] 0 5.81 8.38 10.28 10.64 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:150] "197" "199" "207" "208" ... .. ..$ : chr [1:150] "197" "199" "207" "208" ... $ dmat2 : num [1:150, 1:150] 0 5.09 3.82 7.21 7.59 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:150] "197" "199" "207" "208" ... .. ..$ : chr [1:150] "197" "199" "207" "208" ... $ dmat3 : num [1:150, 1:150] 0 5.63 5.81 8.77 9.14 ... ..- attr(*, "dimnames")=List of 2 .. ..$ : chr [1:150] "197" "199" "207" "208" ... .. ..$ : chr [1:150] "197" "199" "207" "208" ... $ taxa : Factor w/ 50 levels "Streptococcus acidominimus",..: 31 44 26 4 4 31 32 39 42 31 ... $ abbrev: Factor w/ 50 levels "S acidominimus",..: 31 44 26 4 4 31 32 39 42 31 ...
The matrices $dmat1, dmat2, and dmat3 contain
percent nucleotide difference with indels penalized heavily, little, and
somewhat, respectively.
$taxa is a factor of species names; abbreviations of the same
names are found in $abbrev.
data(strep)data(strep)