===== Funções =====
Arquivos para teste:
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:all_genes.csv|}}
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:genes_of_interest.csv|}}
=== match_genes ===
== Código ==
#match_genes finds the "control" genes (gene with different biological function than the gene of interest) closest located to an optimal distance for a given list of genes of interest.
match_genes = function (all, interest, opt = 10^4, graph = F, list=F) {
#Initializes the data.frame that is going to be returned.
match = data.frame(matrix(ncol=3, nrow=0))
#The function is gonna use the median to calculate gene distances.
all$md = apply(all[,c(3,4)], 1, function (x) median(x))
#Same as above, but now for interest data.frame.
interest$md = apply(interest[,c(3,4)], 1, function (x) median(x))
#Here, we start the first loop; we are going to define the control genes one by one.
for (i in 1:nrow(interest)) {
#Subsets the all data.frame containing only genes in the same chromosome as the interest gene.
data = subset(all, all[,2] == interest[i, 2])
#Removes genes that were a match already.
data = subset(data, (!(data[,1] %in% match[,2])))
#Gets only the genes that have different function from the gene of interest. You can use this column as a filter and the algorithm will work the same.
data = subset(data, (data[,5] != interest[i,5]))
#Just in case the data.frame isn't ordered by position in the chromosome. This isn't necessary, but makes it easier to interpret the code when you manually inspect it.
data = data[order(data$md),]
#Calculates the relative distance betwen the gene of interest median position and every gene median position available.
data$dists = abs(interest[i,6] - data[,6])
#Finds the index of the gene that is the closest located to the optimal distance.
j = which.min(abs(data$dists-opt))
#Tests if no gene is found and creates a data.frame with the gene of interest ID, the ID of the gene closest to the optimal distance and the distance itself (these last two may be NAs if no gene is found).
if (!length(j)) {
m = data.frame(interest[i,1], cbind(NA, NA))
} else {
m = data.frame(interest[i,1], data[j,][c(1,7)])
}
#Binds the row created before to the existing match data.frame.
match = rbind(match, m)
}
#Assign names to the columns of match.
colnames(match) = c("interest_ID", "match_ID", "distance")
#Checks if the parameter graph is True.
if (graph) {
#Starts a graphics device X.
x11()
#Plot the distance between genes in a boxplot.
boxplot(match$distance)
#Puts a title to the plot.
title("Distances between each pair gene of interest/control")
#Traces a line in the optimal distance.
abline(h=opt, col="red")
#Closes the graphic device.
dev.off()
}
#Checks what is going to be returned depending on the parameter list
if(list) {
#Returns a list with the match data.frame and the median match distance. It is used in the shuffle_match function.
return (list(match, median(match$distance)))
#If the argument list is false, then only the data.frame match is going to be returned.
} else {
#Returns the match data.frame.
return(match)
}
}
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:match_genes.r}}
== Help ==
match_genes package:misc R Documentation
Description:
The function finds one possible set of control genes, defined as genes
with different biological function (you can code this however you like),
to a given set of genes of interest. The control genes will be defined
based on an optimal distance to the genes of interest.
Usage:
match_genes(all, interest, ...)
Arguments:
all: data.frame containing all genes that could possibly be a control
gene.
interest: data.frame containing the genes of interest.
opt: the optimal distance (i.e., the distance from the genes of interest
to which the control genes must be close to). Default: opt = 10^4
graph: boolean indicating whether a graph with the distances must be
produced or not. Default: graph = F
list: boolean indicating which kind of outcome should be returned: just the data.frame with the matches (list=F) or a list with the data.frame and the median match distance (list=T). Default: list = F
Details:
Both the all data.frame and the interest data.frame must contain five
columns (gene ID, chromosome, start position, end position and the
function). Note that the last column may be treated as a filter, which can
be biological function, gene ontology, etc.
Value:
If list=T:
A list of two elements is returned. The first element is the match data.frame with each interest gene matched to a control gene and the relative distance between these two. The second is the median of the relative distances, which is used by another function.
If list=F:
The match data.frame is returned, which comprises of each interest gene matched to a control gene and the relative distance between these two.
Author(s):
Murillo Fernando Rodrigues
References:
None.
Examples:
#Running match_genes function with both needed data.frames and default parameters. The data.frames used here are available for testing.
all = read.csv ("all_genes.csv", header=T)
interest = read.csv ("genes_of_interest.csv", header=T)
match_genes(all, interest, list = T)
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:help_match_genes.txt}}
----
=== shuffle_match ===
== Código ==
#shuffle_match tests whether changing the order of the rows in the interest data.frame alters the outcome and returns the most optimal result. This function was built because the algorithm of match_genes isn't very smart.
shuffle_match = function (all, interest, opt = 10^4, n = 100) {
#Creates an empty list.
results = list()
#Loop to run match_genes several times.
for (i in 1:n) {
#Shuffle the interest data.frame
interest = interest[sample(nrow(interest)),]
#Appends the results from match_genes to a pre-existing list called results
results = append(results, list(match_genes(all, interest, opt, graph=F, T)))
}
#Gets only the one result which got the median distance closest to the optimal.
results = results[[which.min(abs(sapply(results, function(x) { x[[2]] })-opt))]][[1]]
#Return results
return (results)
}
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:shuffle_match.r}}
== Help ==
shuffle_match package:misc R Documentation
Description:
This function is an improved implementation of match_genes. It uses
brute force to find the most optimal result. shuffle_match runs
match_genes n times and returns the result from match_genes which had
the median of the distances between control and gene of interest closest
to the chosen optimal distance.
Usage:
shuffle_match(all, interest, ...)
Arguments:
all: data.frame containing all genes that could possibly be a control
gene.
interest: data.frame containing the genes of interest.
opt: the optimal distance (i.e., the distance from the genes of interest
to which the control genes must be close to). Default: opt = 10^4
n: number of times to run match_genes. Default: n = 100.
Details:
You should vary n depending on how many genes of interest you have.
Value:
The match data.frame is returned, which comprises of each interest gene
matched to a control gene and the relative distance between these two.
Note that this is the most optimal result the function could find.
Author(s):
Murillo Fernando Rodrigues
References:
None.
Examples:
#match_genes function with both needed data.frames and default parameters.
m = match_genes(all, interest, list = T)
#Check the outcome of match_genes
m
#Running shuffle_match to check whether the order in the interest data.frame alters the outcome. Tipically, doing the shuffle_match is the best way to go, because match_genes algorithm isn't very smart.
s = shuffle_match(all, interest)
#Check the outcome of shuffle_match
s
#Computes the median of the distances between control and interest gene.
md_s = median(r$distance)
#Which result is the closest to the optimal?
(m[[2]]-10^4)
(md_s-10^4)
{{:bie5782:01_curso_atual:alunos:trabalho_final:murillo.rodrigues:help_shuffle_match.txt}}