Overview of loose.rock

André Veríssimo

2021-04-29

Collection of function to improve workflow in survival analysis and data science. Among the many features, the generation of balanced datasets, retrieval of protein coding genes from two public databases (live) and generation of random matrix based on covariance matrix.

The work has been mainly supported by two grants: FCT SFRH/BD/97415/2013 and the EU Commission under SOUND project with contract number 633974.

Install

The only pre-requirement is to install biomaRt bioconductor package as it cannot be installed automatically via CRAN.

All other dependencies should be installed when running the install command.

if (!require("BiocManager"))
  install.packages("BiocManager")
BiocManager::install("loose.rock")

# use the package
library(loose.rock)

Overview

Libraries required for this vignette

library(dplyr)

Get a current list of protein coding genes

Showing only a random sample of 15

coding.genes() %>%
  dplyr::arrange(external_gene_name) %>% {
   dplyr::slice(., sample(seq(nrow(.)), 15)) 
  } %>%
  knitr::kable()
ensembl_gene_id external_gene_name
ENSG00000141469 SLC14A1
ENSG00000267795 SMIM22
ENSG00000134115 CNTN6
ENSG00000186265 BTLA
ENSG00000205208 C4orf46
ENSG00000278167 CCL18
ENSG00000108950 FAM20A
ENSG00000104951 IL4I1
ENSG00000197046 SIGLEC15
ENSG00000151332 MBIP
ENSG00000035499 DEPDC1B
ENSG00000205496 OR51A2
ENSG00000285269 AL160269.1
ENSG00000171747 LGALS4
ENSG00000134758 RNF138

Balanced test/train dataset

This is specially relevant in survival or binary output with few cases of one category that need to be well distributed among test/train data sets or in cross-validation folds.

Example below sets aside 90% of the data to the training set. As samples are already divided in two sets (set1 and set2), it performs the 90% separation for each and then joins (with option join.all = T) the result.

set1 <- c(rep(TRUE, 8), FALSE, rep(TRUE, 9), FALSE, TRUE)
set2 <- !set1
cat(
  'Set1', '\n', set1, '\n\n',
  'Set2', '\n', set2, '\n\n',
  'Training / Test set using logical indices', '\n\n'
)
set.seed(1985)
balanced.train.and.test(set1, set2, train.perc = .9)
#
set1 <- which(set1)
set2 <- which(set2)
cat(
  '##### Same sets but using numeric indices', '\n\n', 
  'Set1', '\n', set1, '\n\n', 
  'Set2', '\n', set2, '\n\n', 
  'Training / Test set using numeric indices', '\n')
set.seed(1985)
balanced.train.and.test(set1, set2, train.perc = .9)
#
#> Set1 
#>  TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE TRUE 
#> 
#>  Set2 
#>  FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE 
#> 
#>  Training / Test set using logical indices 
#> 
#> $train
#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 14 15 17 18 20
#> 
#> $test
#> [1] 13 16 19
#> 
#> ##### Same sets but using numeric indices 
#> 
#>  Set1 
#>  1 2 3 4 5 6 7 8 10 11 12 13 14 15 16 17 18 20 
#> 
#>  Set2 
#>  9 19 
#> 
#>  Training / Test set using numeric indices 
#> $train
#>  [1]  1  2  3  4  5  6  7  8  9 10 11 12 14 15 17 18 20
#> 
#> $test
#> [1] 13 16 19

Generate synthetic matrix with covariance

xdata1 <- gen.synth.xdata(10, 5, .2)
xdata2 <- gen.synth.xdata(10, 5, .75)
#> Using .2^|i-j| to generate co-variance matrix
#> X generated
#>             X1         X2         X3          X4           X5
#> 1   0.58312957 -1.4177825 -1.5962307 -0.36925641  0.727465210
#> 2   0.41677037  1.1936101 -0.1339503 -1.76290605 -0.007014473
#> 3   1.42448267  1.4447604 -0.9234123  0.58587981 -1.055072673
#> 4  -1.05432665 -1.2694366 -0.7239440 -0.40329452 -1.907850014
#> 5   0.08636972 -0.1037405 -0.1262496  0.50288158  1.495560595
#> 6  -0.05170387  0.2428157  1.4867422 -0.65335638 -0.790914639
#> 7   1.26205122 -0.3761988  1.2515329 -0.05340056  0.339303448
#> 8  -0.91086475  1.0808547  0.6094629  1.50616045  0.683470155
#> 9  -0.04990854 -0.6898217  0.7417349  1.34572271 -0.007549189
#> 10 -1.70599973 -0.1050608 -0.5856860 -0.69843065  0.522601581
#> cov(X)
#>       X1    X2   X3    X4     X5
#> 1 1.0000 0.200 0.04 0.008 0.0016
#> 2 0.2000 1.000 0.20 0.040 0.0080
#> 3 0.0400 0.200 1.00 0.200 0.0400
#> 4 0.0080 0.040 0.20 1.000 0.2000
#> 5 0.0016 0.008 0.04 0.200 1.0000

#> Using .75^|i-j| to generate co-variance matrix (plotting correlation)
#> X generated
#>            X1          X2         X3          X4          X5
#> 1  -0.5352761  0.01866608 -0.3626905 -0.79410109 -1.23133292
#> 2  -1.3405531 -1.84315073 -1.2186326 -1.21589142 -0.08130708
#> 3   0.1230853 -0.15684536  0.5826062  1.11404563  0.92246564
#> 4   0.3834488  0.27160718  0.4574595 -0.23430815  0.92046339
#> 5  -1.6754005 -0.13201392  0.1192358  0.69832597  0.37141739
#> 6  -0.2885482 -0.74204784 -1.8912859 -0.90740440 -0.70642514
#> 7   0.2785362 -0.59850263 -0.5850362 -0.74983768 -1.57447837
#> 8   1.4110025  1.00281272  1.0552661  1.92476214  1.47005647
#> 9   1.2933359  1.86340386  0.7103639  0.04042754  0.45349389
#> 10  0.3503693  0.31607064  1.1327139  0.12398146 -0.54435325
#> cov(X)
#>          X1       X2     X3       X4        X5
#> 1 1.0000000 0.750000 0.5625 0.421875 0.3164062
#> 2 0.7500000 1.000000 0.7500 0.562500 0.4218750
#> 3 0.5625000 0.750000 1.0000 0.750000 0.5625000
#> 4 0.4218750 0.562500 0.7500 1.000000 0.7500000
#> 5 0.3164062 0.421875 0.5625 0.750000 1.0000000

Save in cache

Uses a cache to save and retrieve results. The cache is automatically created with the arguments and source code for function, so that if any of those changes, the cache is regenerated.

Caution: Files are not deleted so the cache directory can become rather big.

Set a temporary directory to save all caches (optional)

base.dir(file.path(tempdir(), 'run-cache'))
#> [1] "/tmp/RtmpKnpeCW/run-cache"

Run sum function twice

a <- run.cache(sum, 1, 2)
#> Saving in cache:  /tmp/RtmpKnpeCW/run-cache/8ca6/cache-generic_cache-H_8ca697a81d8184a82de72523a678a4290375a07e304dd20a78bd488827978af3.RData
b <- run.cache(sum, 1, 2)
#> Loading from cache (not calculating):
#>   /tmp/RtmpKnpeCW/run-cache/8ca6/cache-generic_cache-H_8ca697a81d8184a82de72523a678a4290375a07e304dd20a78bd488827978af3.RData
#> Cache was created at 2021-04-29 16:46:43 using loose.rock v1.2.0
all(a == b)
#> [1] TRUE

Run rnorm function with an explicit seed (otherwise it would return the same random number)

a <- run.cache(rnorm, 5, seed = 1985)
#> Saving in cache:  /tmp/RtmpKnpeCW/run-cache/9fda/cache-generic_cache-H_9fdab5baa36653c6d435ce2d68ec6651845f679861f463fe065f38115dc7acbe.RData
b <- run.cache(rnorm, 5, seed = 2000)
#> Saving in cache:  /tmp/RtmpKnpeCW/run-cache/2ada/cache-generic_cache-H_2adac402358921459b509ec972477640ce54df8436844fb57f761cbe49a3296d.RData
all(a == b)
#> [1] FALSE

Proper

One of such is a proper function that capitalizes a string.

x <- "OnE oF sUcH iS a proPer function that capitalizes a string."
proper(x)
#> [1] "One Of Such Is A Proper Function That Capitalizes A String."

Custom colors and symbols

my.colors() and my.symbols() can be used to improve plot readability.

xdata <- -10:10
plot(
  xdata, 1/10 * xdata * xdata + 1, type="l", 
  pch = my.symbols(1), col = my.colors(1), cex = .9,
  xlab = '', ylab = '', ylim = c(0, 20)
)
grid(NULL, NULL, lwd = 2) # grid only in y-direction
for (ix in 2:22) {
  points(
    xdata, 1/10 * xdata * xdata + ix, pch = my.symbols(ix), 
    col = my.colors(ix), cex = .9
  )
}