Podcast: Play in new window | Download
Subscribe: Google Podcasts | Spotify | Stitcher | Email | TuneIn | RSS
As a graduate student, Joel Schwartz developed an immunofluorescence assay for neurotransmitter transport. To quantify his results, he needed to circle the cells in each image so the computer could measure the intensity.
By the time he graduated, Joel had circled over 10 million individual cells.
Over the years, Joel discovered a better way: he taught computers to do the repetitive, complex, and confounding parts of data analysis.
And now he trains other scientists to do the same.
Datum If You’ve Got ‘Em
A typical research lab churns out hundreds of different kinds of data in any given year. There are Western blot bands, gene sequences, patient surveys, micrographs, mass spectra, and myriad other images, spreadsheets and documents.
All of this data must be analyzed, but it’s often a painful process for the researcher. Opening multiple spreadsheets, ‘cleaning up’ missing or erroneous values, and manipulating thousands of rows to form meaningful groups can take hours or days.
We recognize these as repetitive tasks that a computer could do, but not every scientist has the skills to leverage those resources. What is a biologist to do?
The answer may be: ‘learn to code.’ Tech-savvy scientists can use online resources to learn the basics of programming in languages like Python and R. This approach is great because it’s self-paced and can fit within your existing lab schedule. But it can be frustrating when you hit the limits of your understanding, or you try to debug your code without help.
For those who could use a bit more guidance, data analytics bootcamps are great way to race up the learning curve in a few short weeks.
Joel Schwartz, PhD is an instructor at the LEVEL program at Northeastern University, where he provides hands-on experience in data analysis and computer programming.
This week, we ask Joel how learning to code can help bench scientists turn their data into papers even faster. He tells us how he’s using these skills in his own work on autism assessment and how the next big scientific questions may be elucidated by computers.
Cry It Out
As any parent will tell you, kids don’t come with an instruction manual. And if they did, they’d probably have chewed, torn, or burned it to ashes before you got the chance to read it.
Instead, raising a child in the modern age means hearing competing bad advice from other parents and the internet. The sheer volume of terrible ideas can overwhelm even the most earnest of caregivers.
Enter Parentifact.org, the Politifact of… um… parenting facts. We review the great new resource so you can sleep easier at night.
That is, as soon as Jr. stops waking you up every forty-seven minutes.
And if he doesn’t stop screaming, you may want to help yourself to an Octoberfest Beer from Bell’s Brewing in Comstock, MI. It’s only available for a limited time, and just in time for our IPA-free fall. Cheers!
2 thoughts to “081: Data Science Will Accelerate Your Research – with Joel Schwartz, PhD”
I enjoy your discussions immensely! I’m a new listener as I just started my PhD in Oulu, Finland, three days ago and have already started looking at some big time-series datasets using R, which I haven’t used before (newbie :)). I don’t think R will be essential in my program, which is centred around soil and water parameters, but after listening to this episode I feel inclined to try asking my PI if a short course would be allowed/granted as I can see real benefits in general data analysis and visualisation via R. I found an R bootcamp, which is one week and runs in July for 1,500euros.
My question is, how soon is too soon to ask something like this? Remember that I’m new and have only spent three days in the office to date :p
I also want to ask for a week off in August for summer holiday, so I know I need to approach this in a wise and professional fashion.
Any advice via email would be great guys.
Great question! It sounds like you’ve only been in the lab for a little while, and 1,500 euros + time off is a pretty big ask, especially if learning R is not a standard or expected part of your program. My advice is to practice with online learning course for free (e.g. https://www.rstudio.com/online-learning/) and then apply it to your actual research data. Once your PI sees the value of these tools for your research, they’ll be much more likely to spend money on a bootcamp!
Let us know what you do!