Casey Atkins for Nature      
    Aviv Regev likes to work at the edge of what is possible. In    2011, the computational biologist was collaborating with    molecular geneticist Joshua Levin to test a handful of methods    for sequencing RNA. The scientists were aiming to push the    technologies to the brink of failure and see which performed    the best. They processed samples with degraded RNA or    vanishingly small amounts of the molecule. Eventually, Levin    pointed out that they were sequencing less RNA than appears in    a single cell.  
    To Regev, that sounded like an opportunity. The cell is the    basic unit of life and she had long been looking for ways to    explore how complex networks of genes operate in individual    cells, how those networks can differ and, ultimately, how    diverse cell populations work together. The answers to such    questions would reveal, in essence, how complex organisms such    as humans are built. So, we're like, 'OK, time to give it a    try', she says. Regev and Levin, who both work at the Broad    Institute of MIT and Harvard in Cambridge, Massachusetts,    sequenced the RNA of 18 seemingly identical immune cells from    mouse bone marrow, and found that some produced starkly    different patterns of gene expression from the    rest1. They were acting like two    different cell subtypes.  
    That made Regev want to push even further: to use single-cell    sequencing to understand how many different cell types there    are in the human body, where they reside and what they do. Her    lab has gone from looking at 18 cells at a time to sequencing    RNA from hundreds of thousands  and combining single-cell    analyses with genome editing to see what happens when key    regulatory genes are shut down.  
    The results are already widening the spectrum of known cell    types  identifying, for example, two new forms of retinal    neuron2  and Regev is eager to    find more. In late 2016, she helped to launch the International    Human Cell Atlas, an ambitious effort to classify and map all    of the estimated 37 trillion cells in the human body (see    'To build an atlas'). It is part of a    growing interest in characterizing individual cells in many    different ways, says Mathias Uhln, a microbiologist at the    Royal Institute of Technology in Stockholm: I actually think    it's one of the most important life-science projects in    history, probably more important than the human    genome.  
    Such broad involvement in ambitious projects is the norm for    Regev, says Dana Pe'er, a computational biologist at Memorial    Sloan Kettering Cancer Center in New York City, who has known    Regev for 18 years. One of the things that makes Aviv special    is her enormous bandwidth. I've never met a scientist who    thinks so deeply and so innovatively on so many things.  
    When Regev was an undergraduate at Tel Aviv University in    Israel, students had to pick a subject before beginning their    studies. But she didn't want to decide. Too many things were    interesting, she says. Instead, she chose an advanced    interdisciplinary programme that would let her look at lots of    subjects and skip a bachelor's degree, going straight to a    master's.  
    A turning point in her undergraduate years came under the    tutelage of evolutionary biologist Eva Jablonka. Jablonka has    pushed a controversial view of evolution that involves    epigenetic inheritance, and Regev says she admired her courage    and integrity in the face of criticism. There are many easy    paths that you can take, and it's always impressive to see    people who choose alternative roads.  
    Jablonka's class involved solving complicated genetics    problems, which Regev loved. She was drawn to the way in which    genetics relies on abstract reasoning to reach fundamental    scientific conclusions. I got hooked on biology very deeply as    a result, she says. Genes became fascinating, but more so how    they work with each other. And the first vehicle in which they    work with each other is the cell.  
    Regev did a PhD in computational biology under Ehud Shapiro    from the Weizmann Institute of Science in Rehovot, Israel. In    2003 she moved to Harvard University's Bauer Center for    Genomics Research in Cambridge, through a unique programme that    allows researchers to leapfrog the traditional postdoctoral    fellowship and start their own lab. I had my own small group    and was completely independent, she says. That allowed her to    define her own research questions, and she focused on picking    apart genetic networks by looking at the RNA molecules produced    by genes in cells. In 2004, she applied this technique to    tumours and found gene-expression patterns that were shared    across wildly different types of cancer, as well as some that    were more specific, such as a group of genes related to growth    inhibition that is suppressed in acute lymphoblastic    leukaemias3. By 2006, at the age    of 35, she had established her lab at the Broad Institute and    the Massachusetts Institute of Technology in Cambridge.  
    At Broad, Regev continued working on how to tease complex    information out of RNA sequencing data. In 2009, she published    a paper on a type of mouse immune cell called dendritic cells,    revealing the gene networks that control how they respond to    pathogens4. In 2011, she developed    a method that could assemble a complete    transcriptome5  all the RNA being    transcribed from the genes in a sample  without using a    reference genome, important when an organism's genome has not    been sequenced in any great depth.  
    It was around this time that Levin mentioned the prospect of    sequencing the RNA inside a single cell. Up to that point,    single-cell    genomics had been almost impossible, because techniques    weren't sensitive enough to detect the tiny amount of RNA or    DNA inside just one cell. But that began to change around 2011.  
    The study with the 18 immune cells  also dendritic cells  was    meant to test the method. I had kind of insisted that we do an    experiment to prove that when we put the same cell types in,    everything comes out the same, says Rahul Satija, Regev's    postdoc at the time, who is now at the New York Genome Center    in New York City. Instead, he found two very different groups    of cell subtypes. Even within one of the groups, individual    cells varied surprisingly in their expression of regulatory and    immune genes. We saw so much in this one little snapshot,    Regev recalls.  
    I think even right then, Aviv knew, says Satija. When we saw    those results, they pointed the way forward to where all this    was going to go. They could use the diversity revealed by    single-cell genomics to uncover the true range of cell types in    an organism, and find out how they were interacting with each    other.  
    In standard genetic sequencing, DNA or RNA is extracted from a    blend of many cells to produce an average read-out for the    entire population. Regev compares this approach to a fruit    smoothie. The colour and taste hint at what is in it, but a    single blueberry, or even a dozen, can be easily masked by a    carton of strawberries.  
        Reporter Shamini Bundell finds out what can be learned from        studying cells one by one.      
      You may need a more recent browser or to install the latest      version of the Adobe Flash Plugin.    
    By contrast, single-cell-resolved data is like a fruit salad,    Regev says. You can distinguish your blueberries from your    blackberries from your raspberries from your pineapples and so    on. That promised to expose a range of overlooked cellular    variation. Using single-cell genomics to sequence a tumour,    biologists could determine which genes were being expressed by    malignant cells, which by non-malignant cells and which by    blood vessels or immune cells  potentially pointing to better    ways to attack the cancer.  
    The technique holds promise for drug development in many    diseases. Knowing which genes a potential drug affects is more    useful if there's a way to comprehensively check which cells    are actively expressing the gene.  
    Regev was not the only one becoming enamoured with single-cell    analyses on a grand scale. Since at least 2012,     scientists have been toying with the idea of mapping all human    cell types using these techniques. The idea independently    arose in several areas of the world at the same time, says    Stephen Quake, a bioengineer at Stanford University in    California who co-leads the Chan Zuckerberg Biohub. The Biohub,    which has been funding various biomedical research projects    since September 2016, includes its own cell-atlas project.  
    Around 2014, Regev started giving talks and workshops on cell    mapping. Sarah Teichmann, head of cellular genetics at the    Wellcome Trust Sanger Institute in Hinxton, UK, heard about    Regev's interest and last year asked her whether she would like    to collaborate on building an international human cell atlas    project. It would include not just genomics researchers, but    also experts in the physiology of various tissues and organ    systems.  
          I would get stressed out of this world, but she          doesn't.        
    Regev leapt at the chance, and she and Teichmann are now    co-leaders of the Human Cell Atlas. The idea is to sequence the    RNA of every kind of cell in the body, to use those    gene-expression profiles to classify cells into types and    identify new ones, and to map how all those cells and their    molecules are spatially organized.  
    The project also aims to discover and characterize all the    possible cell states in the human body  mature and immature,    exhausted and fully functioning  which will require much more    sequencing. Scientists have assumed that there are about 300    major cell types, but Regev suspects that there are many more    states and subtypes to explore. The retina alone seems to    contain more than 100 subtypes of neuron, Regev says.    Currently, consortium members whose labs are already working on    immune cells, liver and tumours are coming together to    coordinate efforts on these tissues and organs. This is really    early days, says Teichmann.  
    In co-coordinating the Human Cell Atlas project, Regev has    wrangled a committee of 28 people from 5 continents and helped    to organize meetings for more than 500 scientists. I would get    stressed out of this world, but she doesn't, Jablonka says.    It's fun to have a vision that's shared with others, Regev    says, simply.  
    It has been unclear how the project would find funding for all    its ambitions. But in June, the     Chan Zuckerberg Initiative  the philanthropic organization    in Palo Alto, California, that funds the Biohub  contributed    an undisclosed amount of money and software-engineering support    to the Human Cell Atlas data platform, which will be used to    store, analyse and browse project data. Teichmann sees the need    for data curation as a key reason to focus on a large,    centralized effort instead of many smaller ones. The    computational part is at the heart of the project, she says.    Uniform data processing, data browsing and so on: that's a    clear benefit.  
    In April, the Chan Zuckerberg Initiative had also accepted    applications for one-year pilot projects to test and develop    technologies and experimental procedures for the Human Cell    Atlas; it is expected to announce which projects it has    selected for funding some time soon. The applications were open    to everyone, not just scientists who have participated in    planning meetings.  
    Some scientists worry that the atlas will drain both funding    and effort from other creative endeavours      a critique aimed at many such international big-science    projects. There's this tension, says Atray Dixit, a PhD    student in Regev's lab. We know they're going to give us    something, and they're kind of low-risk in that sense. But    they're really expensive. How do we balance that?  
    Developmental biologist Azim Surani at the University of    Cambridge, UK, is not sure that the project will adeptly    balance quantity and depth of information. With the Human Cell    Atlas, you would have a broad picture rather than a deeper    understanding of what the different cell types are and the    relationships between them, he says. What is the pain-to-gain    ratio here?  
    Surani also wonders whether single-cell genomics is ready to    converge on one big project. Has the technology reached    maturity so that you're making the best use of it? he asks.    For example, tissue desegregation  extracting single cells    from tissue without getting a biased sample or damaging the RNA    inside  is still very difficult, and it might be better for    the field, some say, if many groups were to go off in their own    directions to find the best solution to this and other    technical challenges.  
    And there are concerns that the project is practically    limitless in scope. The definition of a cell type is not very    clear, says Uhln, who is director of the Human Protein Atlas     an effort to catalogue proteins in normal and cancerous human    cells that has been running since 2003. There may be a nearly    infinite number of cell types to characterize. Uhln says that    the Human Cell Atlas is important and exciting, but adds: We    need to be very clear, what is the endpoint?  
    Regev argues that completion is not the only goal. It's    modular: you can break this to pieces, she says. Even if you    solve a part of a problem, it's still a meaningful solution.    Even if the project just catalogues all the cells in the    retina, for example, that's still useful for drug development,    she argues. It lends itself to something that can unfold over    time.  
    Regev's focus on the Human Cell Atlas has not distracted her    from her more detailed studies of specific cell types. Last    December, her group was one of three to publish    papers6, 7, 8 in which they used the precision gene-editing    tool CRISPRCas9 to turn off transcription factors and other    regulatory genes in large batches of cells, and then used    single-cell RNA sequencing to observe the effects. Regev's lab    calls its technique Perturb-seq6.  
    The aim is to unpick genetic pathways very precisely, on a much    larger scale than has been possible before, by switching off    one or more genes in each cell, then assaying how they    influence every other gene. This was possible before, for a    handful of genes at a time, but Perturb-seq can work on 1,000    or even 10,000 genes at once. The results can reveal how genes    regulate each other; they can also show the combined effects of    activating or deactivating multiple genes at once, which can't    be predicted from each of the genes alone.  
    Dixit, a co-first author on the paper, says Regev is    indefatigable. She held daily project meetings at 6 a.m. in the    weeks leading up to the submission. I put in this joke    sentence at the end of the supplementary methods  a bunch of    alliteration just to see if anyone would read that far. She    found it, Dixit says. It was 3 a.m. the night before we    submitted.  
    Regev's intensity and focus is accompanied by relentless    positivity. I'm one of the fortunate people who love what they    do, she says. And she still loves cells. No matter how you    look at them, they're just absolutely amazing things.  
Read more:
How to build a human cell atlas : Nature News & Comment - Nature.com