5 Key Takeaways from the Innovations Lab

The following was written by Ellen Eischen, Assistant Professor, Department of Mathematics at the University of Oregon.

A doctor, a mathematician, and a statistician walk into… No, this isn’t the beginning of a joke.  It’s the beginning of the formation of a research team at SAMSI’s five-day Innovations Lab on Interdisciplinary Approaches to Biomedical Data Science Challenges, in which I was fortunate to participate in mid-July at the North Carolina Biotechnology Center.  This innovative NSF-funded pilot program – which brought together 35 experts from mathematics, statistics, computer science, biology, and medicine – facilitated new collaborations in precision medicine.

Participants posted questions of interest.

Precision medicine, which concerns the use of a person’s individual characteristics (e.g. genetic profile, lifestyle, environment) to diagnose, prevent, and treat diseases, is already changing the way medicine is practiced and seems likely to revolutionize treatments, at least for certain classes of diseases.  For example, certain cancer treatments are effective only in patients whose tumors have a particular genetic profile.  Reflecting the urgency for developments in precision medicine, the White House recently announced a major precision medicine initiative.   Projects that began at the SAMSI workshop – which covered a wide range of topics, including pain intervention, psychiatry, mobile technology for public health, and geriatric care – may very well lead toward solutions to medical problems that affect all of us.

Advances in precision medicine rely on analysis of large, complex datasets (commonly referred to as “Big Data”).   Biomedical data tends to be highly heterogeneous – an issue my teams at the workshop repeatedly faced – and thus particularly challenging to handle.  We also repeatedly returned to issues of data quality and accessibility, which in turn led to refining research problems so that they would actually be feasible in the context of currently available datasets.

Structure of the workshop (or How twelve new research groups successfully got off the ground and running in four and a half days)

Due to the unconventional and highly collaborative nature of the workshop, the format differed from a typical workshop.  In collaboration with a team of mentors (experienced scientists who provided feedback), workshop organizers (additional experienced scientists), and officers from the NSF and NIH (who provided crucial guidance concerning funding opportunities), the workshop was partly led and structured by Knowinnovation, a team with vast experience facilitating creative collaborations on interdisciplinary problems.  With Knowinnovation’s website stating “we like to collect people in a room and surprise them with their own ingenuity,” I was skeptical as to whether they would actually have a significant effect in such a short period.  Immediately, though, it was apparent that they were skilled at getting us to interact, share ideas, and turn them into compelling proposals.

The workshop began with an activity designed to get participants to immediately engage with each other and share their expertise and interests.

For most of the first two days, participants met in many different groups, typically for ten to sixty minutes to design a problem, identify the key challenges, and also identify expertise and data needed to address the relevant challenges.  Each session was followed by a brief presentation to the entire group of participants and mentors, and ideas were recorded and posted for all to view throughout the workshop.  At the end of each day, the facilitators from Knowinnovation photographed and posted all these notes to a private online forum, where they also posted photographs and slides from all our presentations, links to databases, and useful articles about collaboration.

Recording sources of important data
Nirmish Shah describes his team’s plans for an app.

Midway through the week, participants formed teams – typically consisting of three to five members –  with whom they would remain for the duration of the workshop.  As we quickly learned from experience while working in groups early in the week, it was usually essential to have a doctor (or other person with medical expertise), someone from mathematics or computer science, and someone with experience in biostatistics in order to make progress.  As an example of the diversity of expertise on a typical team, I will note that my own team consists of a psychiatrist, biostatistician originally trained in mathematics, professor of information sciences specializing in visualization methods and with a background in computer science, computer and electrical engineer, professor of bioinformatics and radiology, and mathematician (me).

Each team spent much of the last two days of the workshop intensely working to refine their ideas in preparation for continued collaboration and for a grant proposal on “Quantitative Approaches to Biomedical Big Data,” due to the NSF’s Division of Mathematical Sciences just two weeks after the workshop ended.  This process was partly aided by helpful feedback.  To start, we spent an afternoon providing feedback to each other’s groups, through a four-stage feedback process consisting of “Pluses, Potentials, Concerns, and Overcoming Concerns” (“PPCO”), which encouraged participants to formulate constructive feedback.  Each group also met with groups of mentors on the last two days to get several rounds of expert feedback, both on the feasibility of the project and on aspects that might need extra work or focus in order to be part of a compelling NSF proposal.

Arianna Di Florio giving a soapbox talk that would soon lead to a new collaboration

Since new ideas about a problem often suddenly arise during the course of thinking about a different problem, there was also time each day for “soapbox talks.”  Anyone could sign up to give a one-minute slide-free talk on a half-baked idea.  Some of the research groups developed in response to ideas proposed in these brief talks.  In fact, the group with whom I ultimately ended up working (and applying for a collaborative NSF grant) initially consisted of several of us with vastly different backgrounds who came together to discuss how to tackle a problem proposed in one of the first soapbox talks.


While most of the week was spent discussing research problems in small groups, the workshop featured several talks (some remote) by experts on precision medicine and data science:

  • Joe Gray, Gordon Moore Endowed Chair in the Biomedical Engineering department and a member of the Knight Cancer Institute at Oregon Health Sciences University, described the use of data science and genomics for cancer treatment.
  • Susan Murphy, a MacArthur fellow and H.E. Robbins Distinguished University Professor of Statistics and Professor of Psychiatry at the University of Michigan, discussed an app her team has designed for personalized health interventions.
  • Bill Noble, a Professor of Genome Sciences and Computer and Electrical Engineering at the University of Washington, spoke about modeling the 4D nucleome (3-dimensional modeling that also accounts for time).
  • DJ Patil, Chief Data Scientist of the United States, gave an engaging talk in which he encouraged the participants to be innovative, emphasizing that “Clever beats smart nine times out of ten” for the sorts of problems we were considering. Patil also highlighted the current administration’s commitment to promoting advances in precision medicine.

Key takeaways

This was one of the most engaging and worthwhile workshops I have attended.  To those who are unfamiliar with this sort of workshop, though, it might sound implausible that thirty-five experts on disparate areas, who had never met before, have since come together to begin functional collaborations and submit compelling grant proposals in such a short period of time.  Since this format of workshop has the potential to lead innovative collaborations in other fields as well, I will conclude by sharing my thoughts on which aspects made the workshop at SAMSI particularly successful and enjoyable:

1. Carefully select participants not just for expertise but also for skills crucial to collaboration.

Richard Smith, the Director of SAMSI, told participants on the first day that the highly selective acceptance procedure (with only 35 accepted participants out of a pool of more than 350 applicants) involved selecting applicants who not only were accomplished in a particular discipline but also were excellent communicators.  In addition to asking about applicants’ professional background, the six-question program application required applicants to discuss their approach to working on teams and their ability to engage and work with non-experts or those with a different perspective.  This led to a highly functional working environment.

2. Involve a team of experienced facilitators.

From accelerating the process of engaging with other participants to approaching feedback effectively (see above), Knowinnovation lived up to its claim of helping “smart people have interesting conversations about complex questions, which leads to novel ideas and innovative research.”  Each person I asked at the end of the workshop said that they found the facilitators from Knowinnovation to have been particularly helpful.

3. Let the structure and schedule be partially participant-driven.

While we received a schedule at the beginning of the workshop, Knowinnovation warned us that it would likely change several times in response to participants’ progress.  This allowed us to work more productively and creatively than a rigid, traditional schedule would.  Also, during the second half of the week, there were large blocks of time open for groups to collaborate without interruption.

4. Have an abundance of mentors readily available.

While progress was participant-driven, mentors provided crucial feedback.  They asked challenging questions that sometimes helped narrow a group’s focus or send it in a more promising direction.

5. Encourage teams to quickly determine and seek whichever expertise they need.

Otherwise, teams get stuck discussing hypotheticals.  Far from the beginning of a joke, it quickly became apparent that the roles “doctor, mathematician, and statistician” from the beginning of this blog post each played an essential role in advancing most of the productive discussions and collaborations from the workshop.

Learning about Neural Spike Train Analysis and More at the SAMSI CCNS Summer School

This week’s blog entry is written by SAMSI’s Kenan Fellow, Alexandra Solender who is the Science Department Chair at Holly Springs High School and teaches AP Physics C, Honors Physics. She is also the Science Olympiad Coach and SNHS Advisor.

Alexandra Solender

Alexandra Solender

During the week of July 27th, SAMSI hosted a Computational Neuroscience Summer School in which I was lucky enough to participate.  The workshop was aimed at researchers in areas such as neuroscience, computer science, applied mathematics, and biomedical engineering as it covered research topics like pattern theory, signal processing, and functional and structural imagining.  This opportunity was extremely exciting for me as I am a high school physics teacher not a math or science researcher.  I have been working with Dr. Tom Witelski at SAMSI since the beginning of June through the Kenan Fellows Program.  The Kenan Fellows Program selects K-12 teachers in North Carolina to work with university or industry experts during the summer to gain real world experience to bring back to their respective classrooms.  As a Kenan Fellow, I worked with Dr. Witelski on bringing data analysis into my classroom.  Because Dr. Witelski was in charge of organizing the Computational Neuroscience workshop I was given the opportunity to sit in on some of the presentations that would be accessible to me given my background and they were all quite interesting!

Tom Witelski and two others at a table

Tom Witelski at the CCNS Summer School.

The first day focused on Neural Spike Train Analysis and Compressed Sensing.  These topics were quite math heavy but the applications were exciting and easier to follow than I was expecting.  The Neural Spike Train Analysis focused on statistical models that capture the structure of the signals in the brain and allow for the prediction of timing of the spikes.  Dr. Uri Eden of Boston University walked us through the processes used to mathematically model physical systems.  Subtopics included poisson distribution (useful models, but don’t fit data well) and Bernoulli Process (a sequence of binary random variables).  The Compressed Sensing talk was led by Justin Romberg of the Georgia Institute of Technology.  His talk focused on signals and images with applications such as radar, MRI scans, sonar, and video cameras.  They all combine pieces/signals to produce a larger picture following an abstract formula of the form y = A x where y is a set of observed data, A is a linear equation system, and x is an unknown.  The compressed sensing was much more complicated but the application was impressive.

Person getting ready to talk

Justin Romberg, Georgia Tech, getting ready to speak at the CCNS Summer School.

The second day of Neural Spike Train Analysis was the talk that related most closely to physics.  Dr. Mark Kramer picked up where his colleague at BU left off.  He talked about biophysical models that simply compared a neural membrane to a simple circuit.  He worked through Ohm’s Law by manipulating the variables based upon what information was most important to the neural spiking.  Dr. Kramer adapted the model for different elements moving through the membrane, making the model more complicated mathematically, but with the intention of better modeling the spikes.  I personally enjoyed this talk the most as it was a topic I had background knowledge in and it was an easy concept to connect biology and physics together.

One of my main goals as an educator is to provide bridges between the sciences so that students can better understand how everything works together.  They so often believe that biology is just biology and physics is just physics when in reality all of the sciences are interrelated.  The partnership between Kenan Fellows and SAMSI has given me the opportunity to see high level applications of my personal educational goals.  It was wonderful to see physics in a new way and get to interact with researchers in the many scientific fields that were represented.