Main Menu (Mobile)- Block

Main Menu - Block

Digging into Data: Q&A with Kristin Branson

custom | custom
custom | custom

06/28/18 | Digging into Data: Q&A with Kristin Branson

node:body | entity_field

As the volume and pace of data collection increase, computation and theory are becoming increasingly vital to biology. How do these research approaches drive science forward at the Janelia Research Campus?

Janelia Senior Group Leader Kristin Branson heads a new core research focus on Computation and Theory. She explains more in this Q&A.

 

Q:

In your own words, what is Computation and Theory (C&T)?

A:

C&T labs work on all aspects of computation that pervade modern biology, from experiment design to data analysis, modeling, and interpretation. We develop machine vision and learning algorithms for analyzing and interpreting the raw data collected at Janelia and elsewhere. We develop computational models and theories to transform these data into an understanding of how and why biological systems operate the way they do. Working with experimental biologists, we design new experiments to refine and test these theories, and fill in the biggest holes in our understanding of biological processes.

While the goals of different C&T labs are quite diverse, one thing that unifies us is our excitement about the new types of experiments being performed at Janelia and figuring out how to gain as much understanding of the world as possible from them. Computer science labs at Janelia are working to automate the analysis and mining of enormous data sets collected through these experiments. Theory labs work with experimental labs to make observations of and manipulate biological processes that were hidden and inaccessible in the past, allowing us to both refine existing models and develop new models beyond the imaginations of past generations.

   

Q:

Why do you think C&T is an important focus area? Why is it important to Janelia?

A:

Computation has become an essential part of biology, and nowhere is this truer than at Janelia. Researchers at Janelia are developing new technologies that produce incredibly large and complex data sets requiring sophisticated computational analyses: for example, fluorescent indicators that make all kinds of biological signals visible; microscopes that can image large, deep volumes with exquisite temporal and spatial resolution; and tools for manipulating biological processes with high precision. The ability to interpret and understand such data at scale is essential to being able to understand complex biological systems such as neural processes that involve the coordinated activity of populations of neurons across the entire brain over extended periods of time. To use these data to their fullest potential requires automated analyses to extract important signal from the raw data, and the complexity of these processes requires the abstractions that theory provides.

   

Q:

Any preliminary goals for C&T?

A:

Our computer science labs want to develop software that enables new types of science and is used by our colleagues at Janelia as well as the general biology community. In collaboration with Mechanistic Cognitive Neuroscience labs, our theory labs want to find out how the nervous system implements new algorithms for cognition. All of us want to develop new paradigms for data-driven theory and collaborations between experimentalists and theorists. An exciting, open question is how to make use of large, complex data sets, such as cellular-resolution neural recordings from large brain regions in behaving animals, to develop altogether new theories that explain how the nervous system in behaving animals might accomplish a particular goal. I’m eager to integrate machine learning, modeling, and theory approaches toward this.

   

Q:

What have you already accomplished in this focus area?

A:

C&T labs have developed computer vision algorithms and software for processing and analyzing large image data sets collected at Janelia. For example, Stephan Saalfeld’s lab has been instrumental in automating the processing of large electron microscopy image volumes of the fly brain. The software developed for these analyses has been freely released and supported, and is being used by many labs around the world. C&T labs have also developed new machine vision and learning algorithms. For example, Srini Turaga’s lab has developed new interpretable deep learning algorithms using variational autoencoders.

Through close, long-term collaborations with experimental labs at Janelia, C&T labs have designed novel experiments and analysis techniques aimed at identifying the algorithms underlying circuit-level computations. This has required a combination of data analysis, data-driven models, and model-driven experiments. For instance, Sandro Romani’s lab, in collaboration with Jeff Magee’s lab, discovered a new type of learning – a novel form of long-term synaptic plasticity in rodent hippocampus. C&T labs have also developed novel theories of brain functions, thanks in part to insights gained from experimental data at Janelia and elsewhere. These theories distill disparate observations into general principles and provide experimentally testable predictions. For example, Sandro’s lab has generalized the knowledge acquired within specific tasks and model systems to develop new theories of memory recall and 3-D spatial navigation, and Ann Hermundstad’s lab has developed new theories of how neural dynamics efficiently support behaviorally relevant computations.

I think one of our greatest successes is our collaborative culture – how closely integrated we are with labs in other focus areas at Janelia. These close collaborations lead to fast progress. Instead of the old feed-forward model – in which biologists collect static data sets, formulate their problems, then ask us for help – we partner with other labs from the beginning. We iteratively and collaboratively design experiments, analysis methods, and theories, jointly considering all these components simultaneously. This allows our computer scientists to work on the most important problems and develop truly useful solutions. And it allows our theorists to design experiments and get the data that best answer biology’s biggest open questions.

   

Q:

Is your group self-driven or driven by other science in the building?

A:

Computer science and computational biology are active areas of scientific research, and our scientists make progress in these academic pursuits. However, all C&T scientists were attracted to Janelia because of the types of experiments being done here and the collaborations that are possible; thus, a large fraction of our work is done in collaboration with other areas at Janelia.

   

Q:

In addition to leading the C&T research area, you’re also managing your lab. How do you think your lab’s work will contribute to C&T research?

A:

My lab develops machine vision and learning systems for quantitatively measuring and analyzing animal behavior, and we endeavor to understand the structure of behavior and how the nervous system produces it. The automated tools we are developing for quantifying behavior are used by our neuroscience colleagues at Janelia as well as by researchers in other areas of biology, including evolutionary biology.

   

Q:

In your lab, what work are you most proud of so far?

A:

I’m proud of our Janelia Automatic Animal Behavior Annotator (JAABA) system, an interactive machine learning system for training automatic animal behavior classifiers. Computer scientist Mayank Kabra and biologist Alice Robie worked together to create a really useful piece of software.

   

Q:

What kinds of people do you think will excel in this new environment at Janelia?

A:

The desire to work closely and communicate effectively with scientists in different areas is essential. It’s also important to be curious and intellectually agile to quickly understand collaborators’ research areas. Having the stubbornness and work ethic to work with real, messy data is important to make real progress in biology. Successful C&T scientists are also creative, open-minded, and eager to try new things.

The research environment at Janelia was designed so that the primary responsibility of researchers is research, and successful C&T scientists will be those who make the best use of this environment. In particular, group leaders have the opportunity to spend a large fraction of their time doing their own research – writing their own programs, proofs, models, and theories.

   

Q:

What would you say to someone who is interested in coming to Janelia?

A:

Janelia is a great place to do science, especially C&T. Because we are internally funded, we have the opportunity to focus on the research questions that we think are most important, including challenging, long-term questions that may not be funded elsewhere. Janelia has a large and growing emphasis on C&T because these fields are essential components of modern biology research.

Our researchers are developing new, creative ways of mixing computation, theory, experiments, and technology development to best answer the biggest questions in biology. Janelia strives not just to make scientific progress, but to make progress in areas that aren’t the standard.