Given the signal and optional control (input) data, determine location of the statistically significant point binding positions. If the control data is not provided, the statistical significance can be assessed based on tag randomization. The method also provides options for masking regions exhibiting strong signals within the control data.
find.binding.positions(signal.data, f=1,e.value = NULL, fdr = NULL, masked.data = NULL,
control.data = NULL, whs = 200, min.dist = 200, window.size = 4e+07, cluster = NULL,
debug = T, n.randomizations = 3, shuffle.window = 1, min.thr = 2, topN = NULL,
tag.count.whs = 100, enrichment.z = 2, method = tag.wtd, tec.filter = T,
tec.window.size = 10000, tec.z = 5, tec.masking.window.size=tec.window.size,
tec.poisson.z=5,tec.poisson.ratio=5, tec = NULL, n.control.samples = 1,
enrichment.scale.down.control =F, enrichment.background.scales = c(1, 5, 10),
use.randomized.controls = F, background.density.scaling = T,
mle.filter = F,min.mle.threshold = 1, ...)signal tag vector list
optional control (input) tag vector list
Fraction of signal read subsampled. Default=1, i.e. no subsampling
E-value defining the desired statistical significance of binding positions.
FDR defining statistical significance of binding positions
instead of determining statistical significance thresholds, return the specified number of highest-scoring positions
window half-sized that should be used for binding detection (e.g. determined from cross-correlation profiles)
optional set of coordinates that should be masked (e.g. known non-unique regions)
minimal distance that must separate detected binding positions. In case multiple binding positions are detected within such distance, the position with the highest score is returned.
size of the window used to segment the chromosome during calculations to reduce memory usage.
optional snow cluster to parallelize the
processing on
debug mode, whether to print debug messages
minimal score requirement for a peak
If TRUE, regions of significant tag enrichment will be masked out when calculating size ratio of the signal to control datasets (to estimate ratio of the background tag density). If FALSE, the dataset ratio will be equal to the ratio of the number of tags in each dataset.
tec
n.control.samples
enrichment.scale.down.control
additional parameters should be the same as those passed
to the lwcc.prediction
number of tag randomziations that should be performed (when the control data is not provided)
Use randomized tag control, even if
control.data is supplied.
during tag randomizations, tags will be split
into groups of shuffle.window and will be maintained
together throughout the randomization.
half-size of a window used to assess fold enrichment of a binding position
Z-score used to define the significance level of the fold-enrichment confidence intervals
In estimating the peak
fold-enrichment confidence intervals, the background tag density is
estimated based on windows with half-sizes of
2*tag.count.whs*enrichment.background.scales.
either tag.wtd for WTD method, or
tag.lwcc for MTC method
If turned on, will exclude predicted positions whose MLE enrichment ratio (for any of the background scales) is below a specified min.mle.threshold
MLE enrichment ratio threshold that each predicted position must exceed if mle.filter is turned on.
Whether to mask out the regions exhibiting significant enrichment in the control data in doing other calculations. The regions are identified using Poisson statistics within sliding windows, either relative to the scaled signal (tec.z), or relative to randomly-distributed expectation (tec.poisson.z).
size of the window used to determine significantly enrichent control regions
size of the window used to mask the area around significantly enrichent control regions
Z-score defining statistical stringency by which a given window is determined to be significantly higher in the input than in the signal, and masked if that is the case.
Z-score defining statistical stringency by which a given window is determined to be significantly higher than the tec.poisson.ratio above the expected uniform input background.
Fold ratio by which input must exceed the level expected from the uniform distribution.
npl A per-chromosome list containing data frames describing determined binding positions. Column description
x: position
y: score
evalue: E-value
fdr: FDR. For peaks higher than the maximum control peak,the highest dataset FDR is reported
enr: lower bound of the fold-enrichment ratio confidence interval. This is the estimate determined using scale of 1. Estimates corresponding to higher scales are returned in other enr columns with scale appearing in the name.
enr.mle: enrichment ratio maximum likely estimate
thr: info on the chosen statistical threshold of the peak scores
# NOT RUN {
# }
# NOT RUN {
# find binding positions using WTD method, 200bp half-window size,
# control data, 1<!-- % FDR -->
bp <-find.binding.positions(signal.data=chip.data,
control.data=input.data,
fdr=0.01,
method=tag.wtd,
whs=200);
# find binding positions using MTC method, using 5 tag randomizations,
# keeping pairs of tag positions together (shuffle.window=2)
bp <- find.binding.positions(signal.data=chip.data,
control.data=input.data,
fdr=0.01,method=tag.lwcc,
whs=200,
use.randomized.controls=T,
n.randomizations=5,
shuffle.window=2)
# print out the number of determined positions
print(paste("detected",sum(unlist(lapply(bp$npl,function(d) length(d$x)))),"peaks"));
# }
Run the code above in your browser using DataLab