Drs. Sriram Pendyala (left) and Doug Fowler:
'There is an inherent difference and distinction to the variant, not differences based on a spectrum.'
A BBI-led paper demonstrates that variant effects span a multi-dimensional continuum, rather than a pathogenic-benign binary, or a one-dimensional spectrum.
The study used VIS-seq, a new image-based method measuring variant effects on molecular and cellular phenotypes in diverse cell types. It expands multiplexed assays of variant effect (MAVEs) beyond simple cell growth, protein abundance, or readouts in cancer cell lines.
“VIS-seq uses a microscope, so we can measure anything a microscope can see,” said BBI’s Doug Fowler, Ph.D., the paper’s corresponding author. “In one experiment, we can see how variants impact what proteins do inside cells, structures like the nucleus, and the cells overall. So, instead of getting one measurement for what a variant does, we get this whole, rich vector of phenotypes altered by the variant. We’re able gain insights into different aspects of variant molecular and cellular function measured simultaneously.”
The paper, “Image-based, pooled phenotyping reveals multidimensional, disease-specific variant effects,” is available on bioRxiv and is currently under review by a scientific journal. In addition to Fowler, several other scientists affiliated with BBI and the Atlas of Variant Effects Alliance contributed to the manuscript, including: Lea Starita, Ph.D., William S. Noble, Ph.D., Alan Rubin, Ph.D., Fritz Roth, Ph.D., Shawn Fayer, Ph.D., Katie Partington, B.S., and Sriram V. Pendyala, Ph.D.
Pendyala, the paper’s first author, noted a key point in the manuscript.
“We purposely chose genes that are implicated in a whole bunch of genetic diseases,” he said. “The first step is to measure more than one thing about each variant. And once you have done that, you plot them in this multi-dimensional space where there is a huge diversity of effects that you can map to disease phenotypes.”
The study applied VIS-seq to approximately 3,000 LMNA and PTEN variants.
The paper states: This “yielded high-dimensional morphological profiles that captured variant-driven changes in protein abundance, localization, activity, and cell architecture… By linking protein variation to cell images at scale, we illuminate how variant effects cascade from molecular to subcellular to cell morphological phenotypes, providing a framework for resolving the complexity of variant function.”
Fowler said the paper explains “an interesting mystery” regarding why some people develop autism. as opposed PTEN Hamartoma Tumor Syndrome, a group of inherited disorders caused by mutations in the PTEN gene.
“We found, using this procedure, that the autism variants – not the tumor syndrome variants – were actually different on a molecular and cellular level,” he said. “The answer is there is an inherent difference and distinction to the variant. Not differences based on a spectrum.”
Despite its promise and potential in studying variants, VIS-seq has “important limitations,” the paper states.
“… Automated image analysis remains a difficult problem,” according to the study. “For example, while image-derived features are useful for morphological profile clustering, these features require manual interpretation. These limitations suggest productive avenues for future work. VIS-seq could be combined with time-lapse imaging or 3D organoid models to capture dynamic or tissue-like phenotypes.”
Pendyala said one of the most important takeaways of the study for other researchers is improved clarity for identifying potential causes of disease.
“One of the things we come away with in this project is when you look at the images, you can put different scales of organization in order,” he said. “And make hypotheses of why a genetic variant might cause a particular phenotype in cell. It’s not only that you have these multiple dimensions of diversity of variant effects, but it’s that the image also teaches you – or at least gives you a guess – of what might be the mechanism behind the disease.”
Fowler said VIS-seq helps address one of the more perplexing issues related to understanding genetic variants: the “why question” and how, at least for now, a question artificial intelligence is unable to answer.
“All AI models do is predict whether a variant is functional or non-functional,” he said. “AI does not tell you why variants are non-functional. But VIS-seq generates this whole vector of phenotypes that answers the “why” question. So, researchers in our lab are now generating more data and using that data to develop models to answer the underlying questions of ‘why.’ That’s the next frontier of variant effects modeling.”
