Zuckerberg Finance: Building Tools for Science Together

The origin of joint computational tools for Human Cell Atlas

Computing Statistics Specialist Kim-Anh La Cao, who works with CZ Biohub scientist Angela Pisco.

Cells are fundamental units of life, but we still have much to learn about their main function and organization. There are thousands of cell types and trillions of individual cells that work in complex systems in order to provide a variety of functions in our body, ranging from the immune system to the brain. New experimental techniques for characterizing individual cells — combined with the right computational approaches — can help us make sense of this complexity and begin to organize it.
')
Human Cell Atlas (HCA) is an ambitious global collaboration to create an open reference map of all cells in the human body through a comprehensive description of cell types, their number and spatial locations. Once completed, it will become a fundamental resource for scientists, allowing them to better understand how healthy cells work, and what happens to them wrong when a disease strikes. But assembling, integrating, analyzing, and sharing this resource requires new cloud-based data infrastructure and new analytical methods for processing and interpreting large and complex different data sets.

CZI supports Human Cell Atlas by providing grants, data infrastructure, collaborative open source software development, and supporting collaborative research. As part of these efforts, CZI Science recently organized a four-day conference of more than 200 scientists, computational biologists and software engineers to initiate the creation of collaborative computing tools for Human Cell Atlas - a series of 85 grants for researchers aimed at working together to solve computational problems for HCA.

Conference Planning for Collaboration

At CZI, we believe that multidisciplinary teams working together accelerate science, especially at the intersection of biology, computing, and software development. But how can we use our scientific conferences to help build cooperation? This meeting provided an opportunity to try out some ideas.

Before the conference, we grouped projects in 12 free research areas organized around data types, analytical methods, and software ecosystems. We named these groups as follows: cell status, cell type, images, multiomics, trajectories, diversity adjustment, abundance variations, potential spaces, compression, scale, ports, and Bioconductor . Some of these groups have actually been used together — others have met each other for the first time.

We asked each group to present their work as a whole, so that we had 12 group presentations, rather than 85 separate projects - and we had lively newsgroups to prepare for the presentations before the meeting. This easy organization and teleconference before the meeting helped defuse the situation when people arrived. We also developed a meeting website with links to repositories, slides, documents, and other projects that were merged into an online center during the meeting.

On the first day, 12 groups presented their projects. After these 12 presentations, the groups then spent the next two and a half days working together to better “share” the work that they could do for one year (the project duration period), emphasizing connecting threads, such as common metrics and controls. data sets for determining the relative success of various algorithms, data standards and metadata standards or integration with various ecosystem programming tools. To add some structure, we also introduced four parallel training sessions to help teach and inform a variety of topics: tools for modern web-based data visualization, the Human Cell Atlas data coordination platform , using and improving bioRxiv for computational biology and collaborative programming. with github .

We spent a lot of time at the meeting so that the groups could work from the time allocated for joint programming to mass brainstorming, diluted with presentations from guests about current major experimental and computational costs associated with HCA - including inspiring basic information from Dana Peer, leader of the computing community HCA.


The meeting ended with the presentations of the participants about what they did and how to continue working together. We have left a lot of unplanned time, both for work and for social interaction, to promote open discussion and building relationships. Choosing a venue that would provide participants with both meeting space and accommodation helped to create a community. Curatorial control datasets in Github can last a year, but smore and karaoke on the beach is a link that can last throughout a scientific career.


Continuation of the discussion with the fire on the ocean.

Thank you @ ciscience for an amazing meeting. It was the most productive, collective and inspiring meeting, which I attended throughout my career!
- Duygu @ duyguucar

What we learned

Three aspects of the meeting were especially inspiring. First, the groups were really happy to collaborate in providing time, space and tools. Secondly, students and academic scientists, along with PI, gave energy to the meeting and probably helped to make sure that the work would actually be done! Third, the introduction of several computational biologists and CZI software engineers helped to facilitate collaboration at the meeting — our team learned a lot from the grantees about the problems in the field, and also helped create inter-sectoral computing capabilities and collaboration.

Everyone was involved in this new structure of meetings in their own way: in the best cases, groups explained the vision and scale of their projects, sometimes literally supervising or defining control data sets or writing down the prototype code. We were also impressed with the groups that found specific opportunities to participate in open-source collaboration with CZI: Josh Moore from the OMERO project now works with the Starfish team on image-based data transcriptome formats, and Ryan Williams and Cotton Seed from the scaling team develop more scalable approaches for storing and calculating matrices in the Data Coordination Platform.

We were also delighted that the developers of the three key software packages for single-core analysis - Scanpy , Seurat and Scater - have made progress in improving the interoperability of their tools and data formats. Just being in the same room, and having time and support, you can stimulate progress.

@DrAnneCarpenter
Chan Zuckerberg Science does amazing things for science.
But working with them was also culturally amazing. This led to scientific discussions with other women experts in computing my age (first for me!). This contributed to the creation of links with different areas, software projects, disciplines. Bravo!
- Ann Carpenter

It was really nice to see that the HCA computing community gathered at this meeting. As one of the participants said to one of us, he felt like “wood fuel of computational science”, and we could not disagree. We are pleased that CZI Science has been able to help give rise to such exciting interactions, and we hope to use this meeting as a model as we continue to launch more and more projects.

To learn more about working in science, visit our website or follow us on Twitter . To learn more about our technology team, subscribe to the CZI technology blog . Sign up for our newsletter to stay up to date on funding opportunities. And you can always contact us at science@chanzuckerberg.com.

Jeremy Freeman, director, specialist in computational biology

Jeremy is a scientist working with the intersection of biology and technology. He wants to understand how biological systems work, and to use this understanding in the interests of both human health and the design of intelligent systems. He studied computer vision at the master's at New York University, headed the neuroscience research lab at Janelia Research Campus HHMI, and is currently at the Chan Zuckerberg Initiative, conducting our work on areas where computing and biology intersect. He is interested in open source and open science, and also unites scientists and engineers in a number of areas.

Arne Bakker, manager of scientific meetings and reviews

Arne Bakker is the leader of scientific meetings in the research group of the Chan Zuckerberg Initiative. He has a doctoral degree in tumor immunology from the Netherlands Cancer Institute, as well as his research at the University of California at Berkeley. Most recently, Arne was an assistant dean in labor education for masters and novice scientists with a doctorate degree at Stanford University. Throughout his career, Arne actively attracted scientists: he was director of the Discovery festival in Amsterdam, jointly organized Beyond Academia at the University of California at Berkeley and PhD Pathways at Stanford, and volunteered for the Bay Area Science Festival. At CZI, Arne combines the lessons he has learned from this multi-faceted career to lead our efforts to unite scientists through meetings, seminars and other meetings with the goal of creating and supporting joint scientific communities.

Translation: Diana Sheremyeva

---