1. Script Metadata for Sharing
Standard Analysis and Code Sharing Working Group
23 January, 2019
Contributor & Presenter:
Bob Friedman, Xybion
Project Lead:
Hanming Tu, Frontage Laboratories

2. Agenda Working Group background Script Metadata for Sharing Project
Issues
Scope
Objectives & Goals
Key Technology and Knowledge
Project Timelines and Tasks

3. Background: Standard Gap Vision: Fill the Gap on Analysis & Displays Standards
Data Collection Systems
Observed Datasets
Analysis Datasets
Tables, Figures and Listings
Clinical Data
Flow
Trial Design
PRM
SDTM
ADaM
No Standards Exist
Industry
Standards
Alignment
CDASH
(Therapeutic Areas)

4. Background: Common Toolbox
Vision: From Everyone Building Their Own Tools to Shared Tools
Text
Shared Value
Shared tools
Paradigm shift: contribution to build a better world
Shared Tools increase the value of the tools

5. Background: Script Repository
Vision: Collaboration platform and Shared Reusable Code Library
Script Repository in Github
White Paper Scripts
Contributed Scripts
Industry
FDA
Academia
Test Data
Script Metadata
SAS, R, Spotfire, Etc.
Code Test Environment

6. Working Group: Suggested New Structure
6 active Projects and many Sub-projects:
Script Repository (4 Sub projects)
Test Data Factory
(TDF, Peter Schaefer)
Analyses and Display White Papers (ADW, Mary Nilsson)
Communication, Promotion, Education (CPE, Jared Slain and Wendy Dobson)
Script Discovery and Acquisition (SDA, Rebeka Revis, Alfredo Rojas)
Repository Content and Delivery (RCD, Gustav Bernard, Andrew Miskell)
Repository Governance and Infrastructure (RGI, Hanming Tu, Mike Carniello)
Script metadata for Sharing (SMS, Hanming Tu)
Best Practices for Quality Control and Validation (Maria Dalton and Jane Marrer)
GPP in Macro Development (Ninan Luke)

7. Background: Scripts Scripts developed by volunteers
6 Script-a-thons (plus additional work by project members) resulting in several scripts at various stages
Scripts developed based on white paper
Scripts contributed by other groups
FDA:
Non-clinical:
Data Handle:
Spotfire Templates:

8. Script Metadata for Sharing
Issue statement:
Not easy to find scripts due to
Script metadata are not defined thus the metadata files are not consistent;
The index page based on metadata files are not updated promptly
Not easy to navigate in the repository due to
Scripts are not well organized;
The folders are deep and complicated
Not easy to use the scripts due to
Need to download the scripts;
Modify the scripts. You usually need to make change to the original script to make it work in your own environment.

9. Project Scope
Definition: Metadata is information about the structures that contain the actual data. A metadata repository is a database used to store, organize, manage, relate/link, and share metadata.
We are NOT to create a MDR system
We are only to define script metadata
Standards:
ISO/IEC (formally known as the ISO/IEC Metadata Registry (MDR) standard) is an international standard for representing metadata for an organization in a metadata registry.
Features:
Generally contains four classifications: ownership, descriptive characteristics, rules and policies, and physical characteristics.
The major categories of structure types are thesaurus, taxonomy, and ontology.
Business cases:
Statistical Computation Environment
Script development life cycle
Define script metadata for Sharing

10. Project Objectives
Script metadata provides the information about the scripts purpose, version, execution environment, library and data files used, inputs, outputs, review history, ratings etc. The metadata will make it easy to share, access and execute scripts in the repository.
Objectives
Clearly define script metadata in a white paper
Conduct a proof of concept (PoC) using the script metadata to drive the sharing and executing scripts in the repository.
Test in both R and SAS scripts

11. Key Technology and Knowledge
Git and Github
YML for Script metadata
Metadata Standards
R, Rstudio, R Packages
SAS

12. Technology Readiness Git and GitHub YML for Script metadata
Have you created a GitHub account?
Have you downloaded a SVN client?
Have you obtained the PhUSE source code
YML for Script metadata
Read about YML?
Read about Metadata Standards?
R, Rstudio, R Packages
Installed R, Rstudio?
Installed DevTools and PhUSE package?
SAS: do you have SAS on your computer?

13. Git and GitHub About Git and GitHub Git Client Github Readme First
Git e-book
GitHub for beginners
Learn: Code Academy quick video tutorial
Git Workflows - "Centralized" is a good starting point
GitHub Workflow - Working with Forks
Git Client
GitHub Clients
Using Git in R/RStudio
Windows: Download TortoiseSVN
Mac: SnailSVN
SmartSVN: download

14. Script Metadata: Format
YML is the chosen format
Keywords: Test, Metadata, Dual Box
Script:
name : metadata_example_rep.yml
title : Metadata example on local drive
desc : > This script demonstrates how to use YML to store the
metadata about your program and define your input parameters
and their values.
version: 0.1.1
Language:
name : YML
version: x.x.x
Environment:
system: Linux or Window 2010
os_version: OEL 5.8, Window 2010
desc: Description of the computing environment.
Inputs:
datasets: dm.xpt,ae.xpt,testfile.xlsx
p1: String - dataset name
p2: Number - depart id
Outputs:
datasets: out1, out2, out3
v1: Date - script execution date and time
v2: User - user who executes the script
Repo:
base_dir: …
lib_files: Func_comm.R
Authors:
- name : Jon Doo
company: PhUSE
Qualification:
last_date: DD-MON-YYYY
last_by: FirstName LastName
stage: T
doc_url: a link to latest documentation
note: C - Contributed; D - Development; T - Testing; Q - Qualified
Stages:
- date: 01-JAN-2016
name: Jon1 Doo
stage: C
docs: a link to qualification documents
Ratings:
- user: htu
date: 25-AUG-2017
asso: Accenture
stars: 5
# end of file
YML is a short name for YAML
Yet Another Markup Language
YAML Ain't Markup Language
YML is a data serialization language that can be read by both human and machine

15. Script Metadata: 8 main Groups
Keywords: used to categorize the script such as analysis, boxplot, etc.
Script: defines the name, version, short and long description of the script.
Language: provides script language info such as SAS 9.4.0, R 3.4.0, etc.
Environment: provides the computing environment of the script language and its configuration.
Inputs: defines the input datasets and parameters required for the successful execution of the scripts. Required.
Parameters: defines the input parameters used by a script. Required.
Outputs: provides the expected output datasets, tables, graphs, etc.
Repo: provides the hosting repository information. Required.

16. Script Metadata: 4 secondary Groups
Authors: documents the developers who create or contribute to the development and qualification of the script.
Qualification: documents the qualification state and process.
Stages: provides the historical status of the scripts.
Ratings: records the users who review and give the rating about the script.

17. R shiny and PhUSE package
Use shiny as interface to develop PhUSE web application framework
R is an open source programming language and software environment for statistical computing and graphics
R Package is the fundamental unit of shareable code bundled with data, tests, examples, and documentation.
RStudio is a free and open-source integrated development environment (IDE) for R.
Rstudio project helps you organize your development and build of a R code or a Package.

18. R shiny and PhUSE package
The PhUSE package sets a web application framework
Shiny is an R package for building interactive web apps straight from R.
The PhUSE package helps finding, downloading and executing scripts.
How to get R PhUSE Package
How to run R PhUSE Package
Or directly install once the package accepted by CRAN
Clone the PhUSE-scripts repository to your local computer if you are the first time to start the interface or the local repository is old
Grep all the YML files from the local repository
Build a data frame to hold the information for all YML files
Write the data frame to a local file
Populate the “Select Script” dropdown list
install.packages("devtools")
library(devtools)
install_github(”TuCai/PhUSE")
library(PhUSE)
start_PhUSE()
install.packages(“PhUSE”)

19. The PhUSE R Package Script: displays the script if it is readable.
YML: displays the content of YML
Info: displays the information about the YML
Metadata: shows the metadata of the script in table format
Verify: verifies the existence of the files defined in YML
Download: downloads the script to local computer
Merge: merges online and local metadata files
Execute: executes the script if it is executable.
# phuse functions:
build_script_df
read_yml
extract_fns
download_fns
cvt_list2df
merge_lists
run_examples
get_inputs

20. PoC: SAS and R examples Whitepaper deliverables
Analyses and Displays Associated with Measures of Central Tendency - Focus on Vital Sign, Electrocardiogram, and Laboratory Analyte Measurements in Phase 2-4 Clinical Trials and Integrated Submission Documents (PDF Version) 10Oct2013
Figure 7-2 Box Plot – Change in xxx Over Time
WPCT
SAS:
Source:
YML: R:
Source:
YML:

21. PoC: R Program and YML R Program YML file library(phuse)
# 1. get the script name and YML name
# the fn() is from phuse web frame work to provide selected YML
le full path and name
yml_fn <- fn();
inYML <- get_inputs(yml_fn)
cfgYML <- read_yml(yml_fn)
# we need to remove any R Shiny specification here
dn <- ifelse(is.null(input$p1), cfgYML$Inputs$p1, input$p1);
nn <- ifelse(is.null(input$p2), cfgYML$Inputs$p2, input$p2);
d <- eval(call(dn, nn))
t <- paste(dn, "(", nn, ")", sep = "")
hist(d, main = t,col="#75AADB", border = "white")
YML file
Inputs:
datasets:
p1: rexp
p2: 100
RShinyUIs :
- control: radioButtons("p1","Distribution type:",c("Normal"="rnorm","Uniform"="runif","Log-normal"="rlnorm","Exponential"="rexp"))
- control: sliderInput("p2","Number of observations:",value = 500,min = 1,max = 1000)

22. PoC: R– more complicated example
R Program
# creation of accompanying datatable
buildtable <- function(avalue, dfname, by1, by2, dignum){
byvarslist <- c(by1,by2)
summary <- eval(dfname)[,list(
n = .N,
mean = round(mean(eval(avalue), na.rm = TRUE), digits=dignum),
sd = round(sd(eval(avalue), na.rm = TRUE), digits=dignum+1),
min = round(min(eval(avalue), na.rm = TRUE), digits=dignum),
q1 = round(quantile(eval(avalue), .25, na.rm = TRUE), digits=dignum),
mediam = round(median(eval(avalue), na.rm = TRUE), digits=dignum),
YML file
Inputs:
datasets: advs.xpt
RShinyUIs :
- control: radioButtons("ft","File type:",c("PNG"="PNG","TIFF"="TIFF","JPEG"="JPEG"))
- control: radioButtons("usn","Use short names?",c("TRUE"="TRUE","FALSE"="FALSE"))
- control: radioButtons("upf","Subset on population flag?",c("TRUE"="TRUE","FALSE"="FALSE"))
- control: radioButtons("trn","Treatment name?",c("TRTA"="TRTA","TRTP"="TRTP"))
- control: sliderInput("perpage","Number per page:",value = 6,min = 1,max = 20)

23. Final Deliverables Script Metadata for Sharing Whitepaper
Public review: done by Dec 13, 2018
PhUSE Web Application Framework:

24. How to participate Sign up for the PhUSE working group mailings
From phusewiki.org, click “Join a Working Group Now”
Standard Scripts Groups
CSS-WG-Standard-Scripts (Entire Working Group)
CSS-WG-Standard-Scripts-WhitePapers (White Paper Project Team)
CSS-WG-Standard-Scripts-Platform (Script Repository)
CSS-WG-SS-WhitePaperReviewers (Notified when a white paper is ready for review)
See wiki pages for each of the projects (

25. Q & A
Contact Information
Hanming Tu, VP of Clinical IT & Database Administration
Phone: ; C:
Bob Friedman ,
Phone: ; C: