In novasmedley/gbmSpm: Find predictive sequential patterns for clinical glioblastoma patients

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

The package gbmSPM can be used to do the exact analysis pipeline used in the paper,

Smedley, Nova F., Benjamin M. Ellingson, Timothy F. Cloughesy, and William Hsu. "Longitudinal Patterns in Clinical and Imaging Measurements Predict Residual Survival in Glioblastoma Patients." Scientific reports 8, no. 1 (2018): 14429.

See also Supplemental Materials.

Vignette Info

This vignette shows how temporal patterns can be found with sequential pattern mining (SPM) using dummy patient data and depends on the arulesSequences R package.

Example

1. Set the arulesSequences parameters

   • minSupp min. percentage of patients, i.e., sequences, with pattern
   • maxgap max. time between CONSECUTIVE elements in sequence, i.e., max. days between visits
   • maxlen max. length of elements in sequence
   • maxsize max. number of items in an element of a sequence

   E.g., if min. support is 0.4, only patterns that occur in 40% of patients will be returned.

   E.g., if max. length is 2, max. gap is 60, patterns are created where there is a sequence of up to 2 visits, with up to 60 days between consecutive visits

   E.g., if max. size is 2, there can be up to two events included in a single visit.

   ```r library(gbmSpm)

   minSupp <- 0.4 maxgap <- 60 maxlen <- 2 maxsize <- 2 ```

2. Set the gbmSPM parameters

   • tType the type of tumor volume measurment to use in generating patterns
   • chemoOverlap whether to only use events during the course of chemotherapy (only really relevant for descriptive statistics)
   • suppList can subset other sets patterns based on level of min. support specified in this list (reduces duplicated work)

   E.g., if tType is 'rate', rate changes will be calculated from tumor volumes and used

   r tType <- 'rate' chemoOverlap <- F suppList <- seq(minSupp, 0.4, .05)

3. Setup experiment to save generated patterns and metadata

   We wil put the created features in '~/gbm_spm_example', but this can be changed.

   r outputDir <- '~/gbm_spm_example' spmDir <- file.path(outputDir,'spm') logDir <- file.path(outputDir,'spm_logs') lapply(c(outputDir, spmDir, logDir), function(i) ifelse(!dir.exists(i), dir.create(i, showWarnings = F, recursive = T), F) )

   Log experiment:

   ```r

   if you want to save the log file, uncomment

   logfn <- file(file.path(logDir,paste0(paste0('g',maxgap,'l',maxlen,'z',maxsize),'_',

   getVolTypeName(tType),'.log')), open='wt')

   sink(logfn, type='output', split=T)

   cat(format(Sys.Date(), format="%Y.%m.%d"), '\n') cat('maxgap: ', maxgap, '\n') cat('maxlen: ', maxlen, '\n') cat('maxsize: ', maxsize, '\n') cat('tumor vol variable type: ', getVolTypeName(tType), '\n\n') ```

4. Load dummy data

   ```r data("fake_data") names(fake_data) colnames(fake_data$person) colnames(fake_data$demo) colnames(fake_data$events)

   save tumor location and laterility strings before event cleaning

   fake_tumorInfo <- fake_data$events fake_demo <- fake_data$demo

   fake_data$events <- cleanData(fake_data$events, tType)

   cat('...',nrow(fake_data$events), " events left for SPM after cleaning", '\n') ```

5. Prep data by collecting patient info for each event

   ```r fake_data <- merge(fake_data$events, fake_data$person, by='iois', all.x=T) colnames(fake_data)

   prep for each event, since age does change

   fake_data <- prepDemographics(fake_data, fake_demo) colnames(fake_data)

   get survival labels, these also change

   fake_data <- prepSurvivalLabels(fake_data) colnames(fake_data)

   get first tumor location

   fake_data <- getTumorLocation(fake_data, fake_tumorInfo) colnames(fake_data)

   ```

6. Run spm

   If you want to save the patterns and use in logit modeling:

   r runSPM(event = fake_data, suppList = suppList, maxgap = maxgap, maxlen = maxlen, maxsize = maxsize, tType = tType, outputDir = spmDir)

   otherwise:

   ```r

   spm

   pSPM <- getSeqPatterns(event = fake_data, transFilename = file.path(outputDir, 'example_transactions.txt'), createT = T, support = minSupp, maxgap = maxgap, maxlen = maxlen, maxsize = maxsize) pSPM$patterns <- as(pSPM$freqseq, "data.frame") pSPM$patterns$sequence <- as.character(pSPM$patterns$sequence)

   days when pattern occur

   patternDays <- findPatternDays(pSPM$patterns, pSPM$data, maxgap=maxgap)

   feature vectors to supply to logits

   feat_vecs <- getFeatureVectors(patternDays, events=pSPM$data) ```

novasmedley/gbmSpm documentation built on May 17, 2019, 10:39 a.m.