In this segment of the Scientist Stories series, Neville Sanjana outlines an ambitious, community‑driven roadmap for mapping the functional consequences of genetic variation. He first describes the goal of creating a comprehensive cis‑regulatory atlas that catalogs regulatory elements and associated variants across many cell types and conditions, coupled with computational models for interpretation. He proposes large‑scale perturbation assays that could eventually test the effect of every nucleotide—coding and non‑coding—on phenotypes such as gene expression, chromatin state, and disease‑relevant traits. Sanjana emphasizes the need to move beyond nearest‑gene assumptions to reliably link causal variants to their target genes, highlighting examples where regulatory variants act at a distance. He discusses the challenges of rapid phenotyping, noting that while cancer models allow straightforward growth or drug‑resistance screens, neurodegenerative contexts like ALS are more difficult. The second major thrust is a gene‑function atlas, leveraging genome‑wide CRISPR knockout, overexpression (ORF), and RNAi libraries to systematically pair each gene with functional readouts, starting with endophenotypes (RNA‑seq, ATAC‑seq, histone modifications) and expanding to disease‑relevant phenotypes. Throughout, Sanjana stresses that these ideas stem from community submissions, aiming to avoid siloed data and to make the resulting resources broadly accessible for the field.
Chapters
Sure. So, I think this is a really nice organization that Jonathan and Nadav put together here, actually, with these different levels of analysis. And so, I think what you'll see here on the slides that we put together is, as Jonathan mentioned, that it's quite general, it's quite open. I think we're going to try and shuttle you through, maybe in just a few minutes, through these few levels and then have most of the time, actually, for discussion. We can try and flesh out the details here. So, the first one here is about cis-regulation. This has been a topic that's come up time and again, I guess, the last two days, linking variants and regulatory elements to their target genes. So, the big goal, the big idea of what we want to achieve is a complete atlas. It's kind of a daring thing, complete atlas of regulatory elements and variants across some number of cell types, some number of conditions. These are the kind of details that can be filled in, and of course, this has to go hand in hand with also developing computational models for interpretation. I think so far the kinds of perturbation assays you see that have been done by many of the folks in this room have been smaller scale and really not given the full computational breadth maybe that they should be. And, of course, you know, the— I think that the nice goal to work toward is perturbation of every single nucleotide in the genome, coding and non-coding, and to look at its effects on phenotypes like gene expression. But also other relevant phenotypes for the model, for the disease, for the trait. Of course, the idea is not just to perturb every single nucleotide, but maybe we can also think about more focused assays looking at sets of variants, trying to identify causal variants out of a confidence set from, say, GWAS, and then trying to say, of— once we identify a variant that seems to have some effect, What gene is it working through? And there's many examples now, elegant examples from folks where it isn't working through the nearest gene. And what kinds— I think something to discuss is what are the, quote, rapid phenotyping screens and platforms, kind of, in some areas like cancer biology, I think it's been very direct and easy. It's growth, it's drug resistance, but then there's been talk today from Tom, from others, of ALS neurodegenerative disease. I think it's not easy there. It's a little bit from experience. Okay, so then the next one is— this is the last one I think that I'm going to speak about, which is what does every gene do? We have these things like genome-wide CRISPR screens. We have had for many years genome-wide overexpression libraries, ORF libraries, RNAi knockdown libraries. So, there's many different ways that we can manipulate every gene in the genome. So, what is the big goal? The big goal is a complete atlas of genes matched with functions, which— what does that mean? Are there an infinite number of functions? Which are the ones to focus on first? And I mentioned this before, that there's endophenotypes, like biochemical phenotypes, like RNA expression, ATAC-seq, histone modifications. These ones, I think, are probably the maybe 1- or 2-year goals where we can pair gene manipulations with these kinds of genomic readouts. But we should also think what other traits, what other disease-relevant phenotypes beyond cancer perhaps even, should we focus on, can we work on? Combination of coding and non-coding. Again, the same thing here, developing rapid phenotypic screens. And I don't know if Jonathan, you mentioned in the beginning that all of these suggestions that we put on these slides are from the ones submitted, right? I'm the reader. These are not my ideas. This is— this came from the submitted set from the entire group. That's— so some of this looks very duplicative, but it's because it's what people said without knowing what everybody else is saying. So just like we were talking about causal variants, dysregulatory variants, the data generation, the analysis has to— the challenge is how do you put together that complete atlas of gene functions? Complete is hard, but some atlas of gene functions that's easy for a large group to access. And it's not the siloed CRISPR screens that I and other people have been responsible for right now. So, okay, hand it off.
