Description Usage Arguments Details Value For 'mean' method For 'loess' method Author(s) See Also Examples

View source: R/normalize_hiC.R

The expected interaction is defined as the linear relationship between the interaction counts and the distance between two loci. See details for additional informations.

1 | ```
getExpectedCounts(x, method=c("mean","loess"), asList=FALSE, ...)
``` |

`x` |
object that inherits from class |

`method` |
method used to estimate the expected counts based on the genomic distance. See details |

`asList` |
return the results as a list. Otherwise, return an
object of class |

`...` |
arguments for mean or loess method, see below |

The expected value is the interaction frequency between two loci that one would expect based on a sole dependency on the genomic proximity of these fragments in the linear genome. This can be estimated using two different methods, mean or loess.

The first method (default) is simply based on the mean counts of each diagonal. If logbin is false, the expected counts will be estimated for each bin of the contact maps. If logbin is true, the binsize will change according to the distance to the diagonal. Short (resp. long) distance will be estimated with smaller (larger) bins. This method works for all resolutions.

The second method is based on a Lowess regression model and works usually fine with low resolution data (250Kb to 1Mb). At higher resolution, the lowess regression might be difficult to fit. The lowess smoothing has two parameters : span and bin. The span corresponds to the fraction of the data used for smoothing. Instead of computing the local polynomial fitting at each data point, a window of size delta (bin parameter) is applied on the data and a linear interpolation is used to fill in the fitted values within the window. The default is 1% of the range of x. If delta=0 all but identical x values are estimated independently. The variance is then estimated using the same span and bin parameter, at each interpolation points. The points inside a window are weighted so that nearby points get the most weight (tricube weight function).

A list with the expected interaction map and the estimated variance

- logbin
if true logarithm based bins are used. In practice, it means that the bin size will change as we move away from the diagonal

- step
multiplicative factor between each bin. Use if logbin is true

- filter.low
fraction of low counts to filter before normalizing. default: 0.05.

- span
fraction of the data used for smoothing at each x point.

- bin
interpolation parameter

- stdev
logical, calculate the variance

- plot
logical, display lowess smoothing and variance estimation points

N. Servant, B. Lajoie

`HTCexp-class`

,`normPerExpected`

,
`normPerExpected`

, `lowess`

1 2 3 4 5 |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.