Package 'FARDEEP'

Title: Fast and Robust Deconvolution of Tumor Infiltrating Lymphocyte from Expression Profiles using Least Trimmed Squares
Description: Using the idea of least trimmed square, it could automatically detects and removes outliers from data before estimating the coefficients. It is a robust machine learning tool which can be applied to gene-expression deconvolution technique. Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie (2019) <doi:10.1101/358366>.
Authors: Yuning Hao [aut], Ming Yan [aut], Blake R. Heath [aut], Yu L. Lei [aut], Yuying Xie [aut, cre]
Maintainer: Yuying Xie <[email protected]>
License: MIT + file LICENSE
Version: 1.0.1
Built: 2025-02-15 05:24:41 UTC
Source: https://github.com/cran/FARDEEP

Help Index

Using the basic idea of least trimmed square to detect and remove outliers before estimating the coefficients. Adaptive least trimmed square.

Description

Using the basic idea of least trimmed square to detect and remove outliers before estimating the coefficients. Adaptive least trimmed square.

Usage

alts(x, y, alpha1 = 0.1, alpha2 = 1.5, k = 6, nn = TRUE,
  intercept = TRUE)

Arguments

x

input matrix of predictors with n rows and p columns.
y

input vector of dependent variable with length n.
alpha1

parameter used to adjust the upper bound of outliers. Take value from 0 to 1, default 0.1.
alpha2

parameter used to adjust the lower bound of outliers. Take value larger than 1, default 1.5.
k

parameter used to determine the boundary of outliers in the following step of algorithm. Take value from 1 to 10, default 6.
nn

whether coefficients are non-negative,default TRUE.
intercept

whether intercept is included in model, default TRUE.

Value

beta: estimation of coefficients.

number_outlier: number of outliers.

outlier_detect: index of detected outliers.

X.new: good observed points for independent variables.

Y.new: good observed points for dependent variables.

k: modified k (if the input value is not appropriate).

Author(s)

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie

References

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie. Fast and Robust Deconvolution of Tumor Infiltrating Lymphocyte from Expression Profiles using Least Trimmed Squares. <doi:10.1101/358366>

Examples

library(FARDEEP)
samp = sample.sim(n = 500, p = 20, sig = 1, a1 = 0.1, a2 = 0.2, nn = TRUE, intercept = TRUE)
result = alts(samp$x, samp$y, alpha1 = 0.1, alpha2 = 1.5, k = 6, nn = TRUE, intercept = TRUE)
coef = result$beta

Using the idea of least trimmed square to detect and remove outliers before estimating the coefficients. A robust method for gene-expression deconvolution.

Description

Using the idea of least trimmed square to detect and remove outliers before estimating the coefficients. A robust method for gene-expression deconvolution.

Usage

fardeep(X, Y, alpha1 = 0.1, alpha2 = 1.5, up = 10, low = 1,
  nn = TRUE, intercept = TRUE, lognorm = TRUE, permn = 100,
  QN = FALSE)

Arguments

X

input matrix of predictors with n rows and p columns.
Y

input vector of dependent variable.
alpha1

parameter used to adjust the upper bound of outliers. Take value from 0 to 1, default 0.1.
alpha2

parameter used to adjust the lower bound of outliers. Take value larger than 1, default 1.5.
up

upper bound of parameter k in function alts, default 10.
low

lower bound of parameter k in function alts, default 1.
nn

whether coefficients are non-negative,default TRUE.
intercept

whether intercept is included in model, default TRUE.
lognorm

whether noise is log-normal distributed, default TRUE.
permn

the number of permutation to get the p-values, default TRUE.
QN

whether perform quantile normalization, default TRUE.

Value

abs.beta: estimation of abosulute abundance of cells (TIL subset scores).

relative.beta: estimation of relative proportions by normalizing abs.beta to 1.

pval: statistical significance for the deconvolution result.

k.value: tuned paprameter by modified BIC.

Author(s)

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie

References

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie. Fast and Robust Deconvolution of Tumor Infiltrating Lymphocyte from Expression Profiles using Least Trimmed Squares. <doi:10.1101/358366>

Examples

library(FARDEEP)
data(LM22)
data(mixture)
# toy examples
result = fardeep(LM22, mixture[, 1:2], permn = 0)

result = fardeep(LM22, mixture)
coef = result$abs.beta

Siganature matrix

Description

A dataset containing 547 genes and 22 TILs.

Usage

LM22

Format

A data frame with 547 rows and 22 variables:

B.cells.naive

naive B cells

B.cells.memory

memory B cells

Plasma.cells

Plasma cells

T.cells.CD8

CD8 T cells

T.cells.CD4.naive

naive CD4 T cells

T.cells.CD4.memory.resting

resting memory CD4 T cells

T.cells.CD4.memory.activated

activated memory CD4 T cells

T.cells.follicular.helper

follicular helper T cells

T.cells.regulatory.Tregs.

regulatory T cells

T.cells.gamma.delta

gamma delta T cells

NK.cells.resting

resting natural killer cells

NK.cells.activated

activated natural killer cells

Monocytes

monocytes

Macrophages.M0

M0 macrophages

Macrophages.M1

M1 macrophages

Macrophages.M2

M2 macrophages

Dendritic.cells.resting

resting dendritic cells

Dendritic.cells.activated

activated dendritic cells

Mast.cells.resting

resting mast cells

Mast.cells.activated

activated mast cells

Eosinophils

eosinophils

Neutrophils

neutrophils

References

Aaron M. Newman, Chih Long Liu, Michael R. Green, Andrew J. Gentles, Weiguo Feng, Yue Xu, Chuong D. Hoang, Maximilian Diehn and Ash A. Alizadeh. Robust enumeration of cell subsets from tissue expression profiles.

Gene-expression data from 14 follicular lymphoma patients

Description

This gene-expression dataset consists of surgical lymph node biopsies of 14 follicular lymphoma patients with 19416 genes. It is available on Gene Expression Omnibus (GEO) with accession number GSE65135. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65135.

Usage

mixture

Format

A data frame with 19416 rows and 14 variables:

GSM1587831

FL lymph node biopsy, untreated, 1063

GSM1587832

FL lymph node biopsy, untreated, 1080

GSM1587833

FL lymph node biopsy, untreated, 575

GSM1587834

FL lymph node biopsy, untreated, 581

GSM1587835

FL lymph node biopsy, untreated, 598

GSM1587836

FL lymph node biopsy, untreated, 639

GSM1587837

FL lymph node biopsy, untreated, 664

GSM1587838

FL lymph node biopsy, untreated, 666

GSM1587839

FL lymph node biopsy, untreated, 695

GSM1587840

FL lymph node biopsy, untreated, 706

GSM1587841

FL lymph node biopsy, untreated, 726

GSM1587842

FL lymph node biopsy, untreated, 731

GSM1587843

FL lymph node biopsy, untreated, 944

GSM1587844

FL lymph node biopsy, untreated, 959

References

Aaron M. Newman, Chih Long Liu, Michael R. Green, Andrew J. Gentles, Weiguo Feng, Yue Xu, Chuong D. Hoang, Maximilian Diehn and Ash A. Alizadeh. Robust enumeration of cell subsets from tissue expression profiles.

Generate random sample with different proportion of outliers and leverage points

Description

Generate random sample with different proportion of outliers and leverage points

Usage

sample.sim(n, p, sig, a1, a2, nn = TRUE, intercept = FALSE)

Arguments

n

number of observations.
p

number of independent variables (predictors).
sig

variance of dependent variable.
a1

proportion of outliers.
a2

proportion of leverage points in outliers.
nn

whether coefficients are non-negative, default TRUE.
intercept

whether intercept is included in model, default TRUE.

Value

y: vector of dependent variable.

x: matrix of predictors with n rows and p columns.

loc: index of added outliers.

beta: vector of coefficients.

Author(s)

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie

References

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie. Fast and Robust Deconvolution of Tumor Infiltrating Lymphocyte from Expression Profiles using Least Trimmed Squares. <doi:10.1101/358366>

Examples

library(FARDEEP)
samp = sample.sim(n = 500, p = 20, sig = 1, a1 = 0.1, a2 = 0.2, nn = TRUE, intercept = TRUE)

Tuning parameter k in function alts using Bayesian Information Criterion (BIC) with some adjustment.

Description

Tuning parameter k in function alts using Bayesian Information Criterion (BIC) with some adjustment.

Usage

tuningBIC(x, y, alpha1 = 0.1, alpha2 = 1.5, up = 10, low = 1,
  nn = TRUE, intercept = TRUE, lognorm = TRUE)

Arguments

x

input matrix of predictors with n rows and p columns.
y

input vector of dependent variable with length n.
alpha1

parameter used to adjust the upper bound of outliers. Take value from 0 to 1, default 0.1.
alpha2

parameter used to adjust the lower bound of outliers. Take value larger than 1, default 1.5.
up

upper bound of parameter k in function alts, default 10.
low

lower bound of parameter k in function alts, default 1.
nn

whether coefficients are non-negative, default TRUE.
intercept

whether intercept is included in model, default TRUE.
lognorm

whether noise is log-normal distributed, default TRUE.

Value

k: tuning result of parameter k for function alts.

Author(s)

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie

References

Yuning Hao, Ming Yan, Blake R. Heath, Yu L. Lei and Yuying Xie. Fast and Robust Deconvolution of Tumor Infiltrating Lymphocyte from Expression Profiles using Least Trimmed Squares. <doi:10.1101/358366>

Examples

library(FARDEEP)
samp = sample.sim(n = 500, p = 20, sig = 1, a1 = 0.1, a2 = 0.2, nn = TRUE, intercept = TRUE)
k = tuningBIC(samp$x, samp$y, lognorm = FALSE)