As a young researcher-in-training at Harvard, chemical biologist Sami Farhi helped develop an innovative technology that can visualize the electrical activity of brain cells. Near the end of his graduate studies in 2018, he lamented the fact that scientists outside his field, especially biologists, often had trouble implementing the tool and interpreting their results. He began envisioning a service that would help users measure brain cell activity in a particular region of a mouse’s brain as the animal completes a behavior, without having to conduct the experiment themselves.
Farhi’s instinct to help more users benefit from new technologies is being put to good use today in the Broad’s Spatial Technology Platform, which he helped establish in late 2022 and now directs. Nearly two dozen members of the platform collaborate with researchers to design experiments that can reveal not only what kinds of cells are present in tissues, but where they are and clues about what they’re doing. These technologies can help answer questions about health and disease using a range of tissue types, animal models, and clinical samples. Platform members generate and analyze data and help users interpret results, in addition to developing and advancing new spatial, imaging, and computational approaches.
We spoke with Farhi about the opportunities made possible by these new spatial approaches and the services he and his team are providing to researchers at the Broad and beyond to make these technologies more accessible.
Why is now the right time to launch this platform?
When I joined the Broad in 2018, scientific leaders at the institute were interested in scaling up our imaging capabilities to support research, particularly those projects aligned with the institute’s Variant to Function initiative. As we built a group focused on optical profiling methods, we watched the field of sequencing-based spatial transcriptomics advance swiftly. Meanwhile, Broad researchers including Fei Chen, Xiao Wang, and Evan Macosko continued to make progress on developing new spatial methods, such as Slide-seq and STARmap, and there clearly was sustained interest. Research in the field began moving beyond simply showing that the methods work to demonstrating real applications in biology.
Over the past few years, launches of new spatial technologies became less frequent as the field solidified a bit, which helped us better keep pace and start to imagine running these protocols at scale, which we’re now doing in the Spatial Technology Platform.
What are some of the promises and challenges of spatial technologies?
Spatial biology approaches combine the virtues of single-cell sequencing with those of more traditional imaging, pathology, and histology approaches to let us investigate tissues in their native context. We can use them to measure the expression of hundreds of proteins or thousands of transcripts at once and map their locations within a tissue or sample. Researchers can use these methods to observe things like how tissues grow and develop and how cells interact with their neighbors, such as how the immune system recognizes a tumor cell or an invading microbe.
But the field is growing fast and it can be difficult for users to keep up. With dozens of new approaches launched in the past several years, it can be intimidating for researchers to choose a technology to use, especially when the startup costs and learning curves are significant. The computational demands can also be incredibly challenging. These datasets are giant. You have to process the data, move it around, visualize it — all of that is difficult. In the Spatial Technology Platform, we’ve built a team of people with profound expertise who can partner with our users to help them extract answers from the raw data. For a more scalable solution, we’re also working on ways to help users help themselves, through sample workflows and guidance.
What spatial technologies is the platform currently supporting?
We aim to provide a range of spatial technologies, both sequencing-based and imaging-based. Today, we run five commercially available spatial platforms including 10X Xenium and Visium, Vizgen MERSCOPE, Curio Seeker, and Lunaphore COMET, and we hope to add more in the near future. Initially we focused exclusively on end-to-end service, starting from the initial consultation through sample processing, data acquisition, and finally through analysis and data delivery back to the user. More recently we began offering training for self-run methods for internal Broad users.
We are also working to become skilled in approaches home-grown at the Broad, so that when those are offered commercially, we are ready to hit the ground running and offer those to our users at scale.
What kinds of projects are underway in the platform?
We support researchers at the Broad, those at external academic research centers, and users in industry who want to use spatial approaches in their work. Some of these projects involve tens of samples, while other partners want to process thousands of clinical samples in dozens of conditions. We also value our relationships with the Broad’s partner hospitals and their pathology and clinical communities, which provide crucial access to patient samples. This sort of work spans cancer oncology, immunology, and neuroscience and makes up the vast majority of our day-to-day operations. Some of the largest efforts are studies aimed at building atlases of the immune system with the ImmGen consortium and of the developing brain with the Brain Initiative Cell Atlas Network, but we’re excited by the prospect for disease-focused studies in the future.
Alongside the platform, I run a research lab where we look for yet-untapped applications of the technology and try to find projects that capitalize on the scale of our operation. Our longest-running collaborations are with Brian Cleary, now at Boston University, and with Ralda Nehme of the Stanley Center for Psychiatric Research here at the Broad. We’re working together to identify cellular mechanisms in neurons that are affected by changes to psychiatric disease risk genes. In the process, we have developed new methods using spatial tools as a screening modality to identify both the genetic perturbation in a cell and the changes to cellular morphology in a single assay. We were the first to use the technology in this way. Some of these genetic variants we’re studying have modest effect sizes, so large-scale data generation efforts like this will become more and more necessary to uncover their impacts on disease.
What’s most exciting about the spatial field today and where it’s headed?
Spatial is such a powerful tool for understanding what makes a large-scale complicated tissue work. It allows us to generate pictures of the full gamut of molecules and where they are and reveal cell-to-cell interactions, all at the same time. It opens up a ton of scientifically interesting questions. One of the most intriguing things about spatial is that we can imagine methods to combine it with examinations of nearly any biomolecule. Given the potential to apply perturbations and read them out in situ (at their original site in a tissue), I imagine that the future will look a little like what we’re seeing in single-cell sequencing now — a shift toward using spatial as a profiling tool for large numbers of experimental conditions, induced by genetic or drug perturbations, but with a much richer profile of the system. For machine learning folks, spatial datasets will be of such a high content and scale that they can power new training models that can help reveal how biological systems work.
Most of all, we’d also love to one day see spatial methods integrate with pathology workflows in the hospital, though we’ll need evidence of the clinical utility and predictive ability of these spatial datasets first. That ability is quite likely there. Considering that spatial combines the utility of both conventional pathology and expression-based readouts, it’s easy to imagine it as an all-inclusive method for profiling patient tissues, providing the full picture of what’s happening in a patient’s tissue in one single snapshot.