another_ds_group_type

Transcript

1. Another type of data science group Leonardo Collado Torres lcolladotor.github.io 2020-08-19
2. LIEBER INSTITUTE for BRAIN DEVELOPMENT What is the role of a data science group? Perform analyses: answer data related questions Stay updated on new methods / software
3. LIEBER INSTITUTE for BRAIN DEVELOPMENT There is increasingly more data & tools - Greater demand for data skills: wrangling, visualization, analysis - LIBD itself generates results that are large data collections - Greater demand across LIBD scientists to learn how to work with data https://ceramics.org/ceramic-tech-today/supercomputer- powered-materials-database-unleashes-data-deluge … and many more
4. LIEBER INSTITUTE for BRAIN DEVELOPMENT What can be the role of a data science group? Perform analyses: answer data related questions Stay updated on new methods / software Provide guidance to others when navigating the world of DS Collaborate as a part of the scientific team - understand the data generation process - work as a team with biologists
5. LIEBER INSTITUTE for BRAIN DEVELOPMENT Proposals for what I want my team to do (you can join us) - Individual data science guidance sessions - General training
6. LIEBER INSTITUTE for BRAIN DEVELOPMENT Proposals for what I want my team to do (you can join us) - Individual data science guidance sessions - Based on my 3 yr experience as an MPH capstone TA - Give ideas and feedback, but don’t do the analysis: different from a consultant or service unit - ~25 min session with ~15 min prep prior and ~10 min post - Build an internal session log (only for those helping) - General training - LIBD rstats club - Public facing knowledge base website
7. LIEBER INSTITUTE for BRAIN DEVELOPMENT Sessions workflow 1. First time: explain your question then meet with me ~25 min 2. Meet with an assigned team member for ~25/50 min 3. Loop back if needed Internally: (session guidance training +) - Slack - Weekly meeting to discuss requests - Access to private session logs Louise Josh Arta Nick You?
8. LIEBER INSTITUTE for BRAIN DEVELOPMENT Some anticipated session topics - Access to & understanding a given dataset - Visualizations: how to make one, alternatives, … - Statistical modeling - Setup & reproducibility in DS: RStudio, Git/GitHub, JHPCE, … - Help learning R/Bioconductor & debugging code - Understanding R objects: SummarizedExperiment, SCE, … - Help with Linux, SGE (JHPCE), software (modules) - Bioinformatics file formats - Data science / bioinformatics concepts & jargon
9. LIEBER INSTITUTE for BRAIN DEVELOPMENT Side effects - To teach, you need to learn (learning has to be included) - You learn differently if you know you’ll teach later - Can identify internal R package / software needs - Break silos: hopefully improve environment - Lead by convincing early on in the life of a project - There is more room for change early on that after a project has been completed
10. LIEBER INSTITUTE for BRAIN DEVELOPMENT Benefits from this group setup & protected time (20%) - Identify projects we want to work on as collaborators - Can help train current group members: level up our DS capacity - Enables delegating more DS tasks & training - Enables building nicer tools/websites/training materials - Reduces bottlenecks by reducing concentration - Provides critical mass to have an effect institute-wide - Can help recruit talent to LIBD & increase exposure - Improve environment &/or productivity?
11. LIEBER INSTITUTE for BRAIN DEVELOPMENT Protected time goes both ways - You need protected time to learn, guide, build training material - You need time also for collaborating - It’s important to respect both and plan accordingly 2 20% 80% research
12. LIEBER INSTITUTE for BRAIN DEVELOPMENT Might help with some of these points raised by Giulio Pergola et al Slide shared with permission
13. LIEBER INSTITUTE for BRAIN DEVELOPMENT With feedback from Louise Huuki Joshua M Stolz Arta Seyedian Nicholas J Eagles Kristen R Maynard Keri Martinowich Stephanie C Page https://media.giphy.com/media/JVXU0uN1l6wdq/giphy.gif