George Church on 4D Omics, Natural Computing, and the BioWeatherMap
By Allison Proffitt
May 4, 2022 | George Church opened the 20th Anniversary Bio-IT World Conference & Expo last night, speaking virtually to a packed room at Boston’s Hynes Convention Center while many more attendees tuned in online.
Church, the Robert Winthrop Professor, Genetics, Harvard Medical School, spoke about both opportunities for IT to serve biology and for biology to serve IT. We have a long way to go on genomics, Church pointed out, with only one telomere-to-telomere genome published, and that not diploid. Beyond human genetics, he highlighted 8.7 million species of eukaryotes along with bacteria, allergens, and viruses. There is so much biology we don’t know or understand yet, he said.
Correlating genotype and phenotype data should be a crucial goal for all of us, he said, mentioning that “there is a program called All of Us, that I think is more recent than the Personal Genome Project.”
Church, of course, founded the Personal Genome Project more than 20 years ago and was one of the first individuals to donate his genome for scientific research. He called for not only genomics and medical record data but all kinds of traits to be gathered. He used the acronym GET—genomes, environments, and traits—and highlighted the infectious viral components of some cancers.
Take a cough for instance, he said, using a pre-pandemic list of possible causes ranging from neural reflex that means nothing to allergies, viruses (including—but not emphasizing—coronaviruses) bacteria and fungi. We really should all be getting very frequent analyses of our environment, Church said, including our food, water, skin microbiome, and more.
The BioWeatherMap project that Church leads with Jason Bobe and Rob Knight is a grassroots effort to do just that. The effort is described as a, “distributed environmental sensing effort aimed at answering some very basic questions about the geographic and temporal distribution patterns of microbial life,” according to the website. “Utilizing the power of high-throughput, low cost DNA sequencing and harnessing the drive of an enlightened public, we propose a new collaborative research approach aimed at generating a steady stream of environmental samples from many geographic locations to produce high quality data for ongoing discovery and surveillance.”
Much of Church’s challenge, in fact, was for an expanded view of what data are important and what we might be able to accomplish with them. He challenged the Bio-IT World audience to expand their thinking of ‘omics beyond just reading and writing genomes to include RNA, nucleic acids, DNA and both sequencing and quantifying. Almost every kind of ‘omics can be done in situ, he said, in both living and fixed cells. He also challenged researchers to look beyond 3D—tracking cells at both base pair and spatial resolutions—to 4D studies, looking at pathologies over time.
In mouse models, Church’s lab has been using CRISPR to track 4D physiological development. He inserted self-guiding RNA into nearly all chromosomes of a mouse. “Each of them during development, or during a time series on a physiological measurement, will be changing itself and thereby recording time series data.”
You’d think this would cause some sort of havoc like cancers, he said, but the animals make it to adulthood and are “relatively healthy.” Such mice, he said, are distributed by the NIH Mutant Mouse Resource Center.
Early attempts at this have added about one billionth of the mass of the mouse to the mouse genome, but Church envisions increasing this percentage, targeting the repetitive elements of the genome. These sections are not “junk DNA” Church was quick to point out—referencing his own graduate thesis—but they are areas that it may be “still be safe to write into.”
This refined process is not “genome vandalism” as he called CRISPR, but a finer process that swaps cytosine for thymine and adenine for guanine, and has avoided nicking and double stranded breaks, he reported.
On the IT side, Church explored how we can use nature as a model for natural computing that would take forever on typical silicon chips. He highlighted recent work out of his lab that used machine learning and “mega libraries” to synthesize things we’d otherwise just simulate. The team published a unified rational protein engineering report in Nature Methods in late 2019. The work has since given rise to several different companies including Dino Therapeutics, Manifold Bio, and Patch Bioscience.
The technical loop, Church explains, starts with machine learning, but with minimal input data. It’s then augmented by DNA synthesis, selection, and DNA sequencing. We end up with a library of designed viruses comprehensive of all the options, he says. Adding logical regression, convolutional neural networks, or recurrent neural networks delivers viable viruses with all 28 amino acids changed, useful—he pointed out—for several applications include patent avoidance.
When asked by the audience what he really wishes would shake up the space, Church chose to highlight public participation from an apparently long wish list. He mentioned how individuals have contributed to—and been respected by—astronomy and meteorology. He hopes for the same “citizen science attitude” in genome sequencing and data collection and sharing in the biology space—again plugging the efforts of BioWeatherMap.
“You’re seeing things like this bubble up in projects around the world. It would be nice to get to some billions of us sequenced and combined with comprehensive trait data and environmental data.”