| Title: | Algorithm for Ranking Predictors Using Graphical Domain Knowledge |
|---|---|
| Description: | Grows families of features by selecting features that maximize a weighted score calculated from empirical feature scores and graphical knowledge. The final weighted score for a feature is determined by summing a feature's family-weighted scores across all families in which the feature appears. |
| Authors: | Michelle Saul |
| Maintainer: | Michelle Saul <[email protected]> |
| License: | GPL |
| Version: | 1.0 |
| Built: | 2024-11-20 06:25:30 UTC |
| Source: | CRAN |
Grows families of features by selecting features that maximize a weighted score calculated from empirical feature scores and graphical knowledge. The final weighted score for a feature is determined by summing a feature's family-weighted scores across all families in which the feature appears.
The DESCRIPTION file:
| Package: | FamilyRank |
| Type: | Package |
| Title: | Algorithm for Ranking Predictors Using Graphical Domain Knowledge |
| Version: | 1.0 |
| Date: | 2021-01-24 |
| Author: | Michelle Saul |
| Maintainer: | Michelle Saul <[email protected]> |
| Description: | Grows families of features by selecting features that maximize a weighted score calculated from empirical feature scores and graphical knowledge. The final weighted score for a feature is determined by summing a feature's family-weighted scores across all families in which the feature appears. |
| License: | GPL |
| Imports: | Rcpp (>= 1.0.6), plyr (>= 1.8.6), stats (>= 3.6.0) |
| LinkingTo: | Rcpp, RcppArmadillo |
| LazyData: | true |
| NeedsCompilation: | yes |
| Packaged: | 2021-01-29 21:06:01 UTC; mwinerip |
| Repository: | CRAN |
| Date/Publication: | 2021-02-05 08:50:08 UTC |
Index of help topics:
FamilyRank-package Algorithm for Ranking Predictors Using
Graphical Domain Knowledge
createCase Simulate Cases
createControl Simulate Control
createData Simulate Data
createGraph Simulate Graph
familyRank Feature Ranking with Family Rank
grow Grow Families
indexFeats Re-index features
rbinorm Bimodal Normal Distribution
The main function is familyRank.
Michelle Saul
Maintainer: Michelle Saul <[email protected]>
ADD REFERENCE
Numerical feature simulation for positive samples. Called by createData.
createCase(subtype, upper.mean, lower.mean, upper.sd, lower.sd, n.features, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)createCase(subtype, upper.mean, lower.mean, upper.sd, lower.sd, n.features, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)
subtype |
Numeric number indicating which subtype to simulate. Currently supports three subtype: 1, 2 or 3. |
upper.mean |
The mean of the upper component of the bimodal Gaussian distribution from which features are simulated. |
lower.mean |
The mean of the lower component of the bimodal Gaussian distribution from which features are simulated. |
upper.sd |
The standard deviation of the upper component of the bimodal Gaussian distribution from which features are simulated. |
lower.sd |
The standard deviation of the lower component of the bimodal Gaussian distribution from which features are simulated. |
n.features |
Number of features to simulate. |
subtype1.feats |
Numeric vector representing the indices of features that define subtype 1. |
subtype2.feats |
Numeric vector representing the indices of features that define subtype 2. |
subtype3.feats |
Numeric vector representing the indices of features that define subtype 3. |
Simulations support 3 subtypes, each defined by 5 different features.
Subtype 1 is defined as having the first 3 subtype1.feats and at least one of the next 2 subtype1.feats simulated from the upper component of the bimodal Gaussian distribution.
Subtype 2 is defined as having all 5 subtype2.feats simulated from the upper component.
Subtype 3 is defined as having the first 4 subtype3.feats simulated from the upper component and and the last subtype3.feats simulated from the lower component.
Returns a vector of simulated features
createCase is not meant to be called alone. It is designed as a helper function for createData.
Michelle Saul
ADD REFERENCE
# Toy Example case <- createCase(subtype = 1, upper.mean = 13, lower.mean = 5, upper.sd = 1, lower.sd = 1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)# Toy Example case <- createCase(subtype = 1, upper.mean = 13, lower.mean = 5, upper.sd = 1, lower.sd = 1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)
Numerical feature simulation for negative samples. Called by createData.
createControl(upper.mean, lower.mean, upper.sd, lower.sd, n.features, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)createControl(upper.mean, lower.mean, upper.sd, lower.sd, n.features, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)
upper.mean |
The mean of the upper component of the bimodal Gaussian distribution from which features are simulated. |
lower.mean |
The mean of the lower component of the bimodal Gaussian distribution from which features are simulated. |
upper.sd |
The standard deviation of the upper component of the bimodal Gaussian distribution from which features are simulated. |
lower.sd |
The standard deviation of the lower component of the bimodal Gaussian distribution from which features are simulated. |
n.features |
Number of features to simulate. |
subtype1.feats |
Numeric vector representing the indices of features that define subtype 1. |
subtype2.feats |
Numeric vector representing the indices of features that define subtype 2. |
subtype3.feats |
Numeric vector representing the indices of features that define subtype 3. |
Simulates data such that none of the 3 subtypes defined in createCase are represented.
To ensure subtype 1 is not represented, at least one of the first three subtype1.feats and/or both of the next 2 subtype1.feats are simulated from the lower component of the Gaussian distribution.
To ensure subtype 2 is not represented, at least one of the five subtype2.feats is simulated from the lower component.
To ensure subtype 3 is not represented, at least one of the first 4 subtype3.feats is simulated from the lower component and/or the last subtype3.feats is simulated from the upper component.
Returns a vector of simulated features
createControl is not meant to be called alone. It is designed as a helper function for createData.
Michelle Saul
ADD REFERENCE
# Toy Example control <- createControl(upper.mean = 13, lower.mean = 5, upper.sd = 1, lower.sd = 1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)# Toy Example control <- createControl(upper.mean = 13, lower.mean = 5, upper.sd = 1, lower.sd = 1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)
Simulate data sets meant to emulate gene expression data in oncology.
createData(n.case, n.control, mean.upper = 13, mean.lower = 5, sd.upper = 1, sd.lower = 1, n.features = 10000, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)createData(n.case, n.control, mean.upper = 13, mean.lower = 5, sd.upper = 1, sd.lower = 1, n.features = 10000, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15)
n.case |
Number of cases to simulate. |
n.control |
Number of controls to simulate |
mean.upper |
Mean of upper component of bimodal Gaussian distribution from which features are simulated. |
mean.lower |
Mean of lower component of bimodal Gaussian distribution from which features are simulated. |
sd.upper |
Standard deviation of upper component of bimodal Gaussian distribution from which features are simulated. |
sd.lower |
Standard deviation of lower component of bimodal Gaussian distribution from which features are simulated. |
n.features |
Number of features to simulate |
subtype1.feats |
Index of features used to define subtype 1. |
subtype2.feats |
Index of features used to define subtype 2. |
subtype3.feats |
Index of features used to define subtype 3. |
Simulates case/control data as described in createCase and createControl, and graphical domain knowledge as described in createGraph.
Returns a named list with a simulated feature matrix (x), simulated binary response vector (y), vector of subtype labels (subtype), and simulated domain knowledge graph (graph).
Michelle Saul
ADD REFERENCE
createCase, createControl, createGraph
## Toy Example # Simulate data set # 10 samples # 20 features # Features 1 through 15 perfectly define response # All other features are random noise. data <- createData(n.case = 5, n.control = 5, mean.upper=13, mean.lower=5, sd.upper=1, sd.lower=1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15) x <- data$x y <- data$y graph <- data$graph## Toy Example # Simulate data set # 10 samples # 20 features # Features 1 through 15 perfectly define response # All other features are random noise. data <- createData(n.case = 5, n.control = 5, mean.upper=13, mean.lower=5, sd.upper=1, sd.lower=1, n.features = 20, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15) x <- data$x y <- data$y graph <- data$graph
Simulate domain knowledge graph.
createGraph(subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15, n.interactions = 1e+06, n.features = 10000)createGraph(subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15, n.interactions = 1e+06, n.features = 10000)
subtype1.feats |
Index of features used to define subtype 1. |
subtype2.feats |
Index of features used to define subtype 2. |
subtype3.feats |
Index of features used to define subtype 3. |
n.interactions |
Number of pairwise interactions to simulate. |
n.features |
Number of features to simulate |
Returns a data frame representation of a graph. The first two columns represent graph nodes and the third column represents the edge weights between nodes.
All pairwise combinations of subtype1.feats have an edge weight of 1.
All pairwise combinations of subtype2.feats have an edge weight of 1.
All pairwise combinations of subtype3.feats have an edge weight of 1.
All other pairwise combinations have an edge weight uniformly distributed between 0 and 1.
Michelle Saul
ADD REFERENCE
# Toy Example graph <- createGraph(subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15, n.interactions = 100, n.features = 20)# Toy Example graph <- createGraph(subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15, n.interactions = 100, n.features = 20)
Ranks features by incorporating graphical knowledge to weight empirical feature scores. This is the main function of the FamilyRank package.
familyRank(scores, graph, d = 0.5, n.rank = min(length(scores), 1000), n.families = min(n.rank, 1000), tol = 0.001)familyRank(scores, graph, d = 0.5, n.rank = min(length(scores), 1000), n.families = min(n.rank, 1000), tol = 0.001)
scores |
A numeric vector of empirical feature scores. Higher scores should indicate a more predictive feature. |
graph |
A matrix or data frame representation of a graph object. |
d |
Damping factor |
n.rank |
Number of features to rank. |
n.families |
Number of families to grow. |
tol |
Tolerance |
The scores vector should be generated using an existing statistical method. Higher scores should correspond to more predictive features. It is up to the user to adjust accordingly. For example, if the user wishes to use p-values as the empirical score, the user should first adjust the p-values, perhaps by subtracting all p-values from 1, so that a higher value corresponds to a more predictive feature.
The graph must be supplied in matrix form, where the first two columns represent graph nodes and the third column represents the edge weights between nodes. The graph nodes must be represented by the index of the feature that corresponds with the index in the score vector. For example, a node corresponding to the first value of the score vector should be indicated by a 1 in the graph object, the second by a 2, etc. It is not necessary that every feature in the score vector appear in the graph. Missing pairwise interactions will be considered to have interaction scores of 0.
The damping factor, d, represents the percentage of weight given to the interaction scores. The damping factor must be between 0 and 1. Higher values give more weight to the interaction score while lower values give more weight to the empirical score.
The value for n.rank must be less than or equal to the number of scored features. The algorithm will include only the top n.rank features in the ranking process (e.g. the n.rank features with the highest values in the score vector will be used to grow families). Higher values of n.rank require longer compute times.
The value for n.families must be less than or equal to the value of n.rank. This is the number of families the algorithm will grow. If n.families is less than n.rank, the algorithm will initate families using the n.families highest scoring features. Higher values of n.families require longer compute times.
The tolerance variable, tol, tells the algorithm when to stop growing a family. Features are added to families until the weighted score is less than the tolerance level, or until all features have been added.
Returns a vector of the weighted feature scores.
Michelle Saul
ADD REFERENCE
# Toy Example scores <- c(.6, .2, .9) graph <- cbind(c(1,1), c(2,3), c(.4, .8)) familyRank(scores = scores, graph = graph, d = .5) # Simulate data set # 100 samples # 1000 features # Features 1 through 15 perfectly define response # All other features are random noise simulatedData <- createData(n.case = 50, n.control = 50, mean.upper=13, mean.lower=5, sd.upper=1, sd.lower=1, n.features = 10000, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15) x <- simulatedData$x y <- simulatedData$y graph <- simulatedData$graph # Score simulated features using absolute difference in group means scores <- apply(x, 2, function(col){ splt <- split(col, y) group.means <- unlist(lapply(splt, mean)) score <- abs(diff(group.means)) names(score) <- NULL return(score) }) # Display top 15 features using emprical score order(scores, decreasing = TRUE)[1:15] # Rank scores using familyRank scores.fr <- familyRank(scores = scores, graph = graph, d = .5) # Display top 15 features using emprical scores with Family Rank order(scores.fr, decreasing = TRUE)[1:15]# Toy Example scores <- c(.6, .2, .9) graph <- cbind(c(1,1), c(2,3), c(.4, .8)) familyRank(scores = scores, graph = graph, d = .5) # Simulate data set # 100 samples # 1000 features # Features 1 through 15 perfectly define response # All other features are random noise simulatedData <- createData(n.case = 50, n.control = 50, mean.upper=13, mean.lower=5, sd.upper=1, sd.lower=1, n.features = 10000, subtype1.feats = 1:5, subtype2.feats = 6:10, subtype3.feats = 11:15) x <- simulatedData$x y <- simulatedData$y graph <- simulatedData$graph # Score simulated features using absolute difference in group means scores <- apply(x, 2, function(col){ splt <- split(col, y) group.means <- unlist(lapply(splt, mean)) score <- abs(diff(group.means)) names(score) <- NULL return(score) }) # Display top 15 features using emprical score order(scores, decreasing = TRUE)[1:15] # Rank scores using familyRank scores.fr <- familyRank(scores = scores, graph = graph, d = .5) # Display top 15 features using emprical scores with Family Rank order(scores.fr, decreasing = TRUE)[1:15]
Call to the C++ function that grows the families.
grow(n, f, d, graph, scores, feat_mat, score_mat, tol, weight_mat, selected)grow(n, f, d, graph, scores, feat_mat, score_mat, tol, weight_mat, selected)
n |
Number of features to rank. |
f |
Number of families to grow. |
d |
Damping factor |
graph |
A matrix or data frame representation of a graph object. |
scores |
A numeric vector of empirical feature scores. |
feat_mat |
Matrix to store selected features. |
score_mat |
Matrix to store weighted scores of selected features. |
tol |
Tolerance |
weight_mat |
A matrix to store the cumulative weighted scores of selected futures across all families. |
selected |
Vector indicating whether a feature has been selected yet. |
This is the workhorse function for the Family Rank algorithm.
Returns a matrix with 1+2xn.families columns and n.rank rows. The first column is the cumulative feature score for each of the ranked features 1:n.rank. The row number corresponds to the re-indexed feature index. The next n.families columns contain the indices of selected features for each iteration of feature selection. The last n.families columns contain the weigthed scores of selected features for each iteration.
Michelle Saul
ADD REFERENCE
# Toy Example scores <- c(.6, .2, .9) graph <- cbind(c(1,1), c(2,3), c(.4, .8)) # initialize matrices n <- n.families <- length(scores) feat.mat <- score.mat <- matrix(0, nrow = n, ncol = n.families) feat.mat[1,] <- order(scores, decreasing = TRUE) score.mat[1,] <- sort(scores, decreasing = TRUE) # Grow families mats <- grow(n = n, f = n.families, d = 0.5, graph = as.matrix(graph), scores = scores, feat_mat = feat.mat, score_mat = score.mat, tol = 0, weight_mat = as.matrix(scores), selected = rep(1, n)) # Selected Feature Matrix ## columns represent familes ## rows represent iterations ## values indicate indices of selected features feat.mat <- mats[, 2:(n.families+1)] feat.mat # Corresponding Score Matrix ## columns represent familes ## rows represent iterations ## values indicate max weighted score of selected features score.mat <- mats[, (n.families+2):(1+2*n.families)] score.mat# Toy Example scores <- c(.6, .2, .9) graph <- cbind(c(1,1), c(2,3), c(.4, .8)) # initialize matrices n <- n.families <- length(scores) feat.mat <- score.mat <- matrix(0, nrow = n, ncol = n.families) feat.mat[1,] <- order(scores, decreasing = TRUE) score.mat[1,] <- sort(scores, decreasing = TRUE) # Grow families mats <- grow(n = n, f = n.families, d = 0.5, graph = as.matrix(graph), scores = scores, feat_mat = feat.mat, score_mat = score.mat, tol = 0, weight_mat = as.matrix(scores), selected = rep(1, n)) # Selected Feature Matrix ## columns represent familes ## rows represent iterations ## values indicate indices of selected features feat.mat <- mats[, 2:(n.families+1)] feat.mat # Corresponding Score Matrix ## columns represent familes ## rows represent iterations ## values indicate max weighted score of selected features score.mat <- mats[, (n.families+2):(1+2*n.families)] score.mat
Re-index features based on number to rank. Called by familyRank.
indexFeats(scores, graph, n.rank = NULL)indexFeats(scores, graph, n.rank = NULL)
scores |
A numeric vector of empirical feature scores. |
graph |
A matrix or data frame representation of a graph object. |
n.rank |
Number of features to rank. |
This function is used to re-index features for the Family Rank algorithm. The function takes in the scores for all features, and returns scores for the top n.rank features. It also takes in the full domain knowledge graph and returns the subgraph that only includes interactions between the top n.rank features. Finally, it re-indexes the top features in both the score vector and domain knowledge graph to 1:n.rank.
Returns a named list with re-indexed domain knowledge graph (graph.w), re-indexed scores (score.w), a mapping between original and new indices (loc.map), and the number of features to rank (n.rnak).
indexFeats is not meant to be called alone. It is designed as a helper function for familyRank.
Michelle Saul
ADD REFERENCE
Simulates random data from a bimodal Gaussian distribution.
rbinorm(n, mean1, mean2, sd1, sd2, prop)rbinorm(n, mean1, mean2, sd1, sd2, prop)
n |
Number of observations to simulate |
mean1 |
Mean of mode 1 |
mean2 |
Mean of mode 2 |
sd1 |
Standard deviation of mode 1 |
sd2 |
Standard deviation of mode 2 |
prop |
Probability of being in mode 1. |
This function is modeled off of the rnorm function.
Generates random deviates
Michelle Saul
## Generate 100 samples from a two component Guassian curve samples <- rbinorm(n=100, mean1=10, mean2=20, sd1=1, sd2=2, prop=.5) ## Plot distribution of simulated data plot(density(samples))## Generate 100 samples from a two component Guassian curve samples <- rbinorm(n=100, mean1=10, mean2=20, sd1=1, sd2=2, prop=.5) ## Plot distribution of simulated data plot(density(samples))