feature Voices in methods development To mark the 15th anniversary of Nature Methods, we asked scientists from across diverse felds of basic biology research for their views on the most exciting and essential methodological challenges that their communities are poised to tackle in the near future.

Polina Anikeeva: Clifford Oliver Fiehn: Neural engineering Brangwynne: We Metabolomics has benefitted have a detailed has become an from decades of understanding of integral cornerstone innovation in micro- the conditions under of biological and nano-electronics, which distinct states research. Biological photonics, materials of non-living matter interpretations science, chemistry form, codified in rely on accurate and synthetic biology. phase diagrams that identification Our current ability to reflect underlying of metabolites. Credit: Andrew integrate these fields thermodynamic driving forces. Can we Yet, currently, compound annotations LaNoue with each other and achieve a similar quantitative understanding lack confidence scoring; this needs to with neuroscience, of liquid–liquid phase separation within change! Data reports should become however, pales in comparison with the living cells? To truly understand intracellular more harmonized, with cloud processing scale and complexity of neuronal signaling. self-assembly, and its functional and for large data sets and kits of internal Understanding the nervous system in the pathological dysregulation in devastating standards to assess metabolite levels. Even context of health and disease will demand diseases, the answer needs to be yes. New in-depth untargeted discovery assays should a paradigm shift, from refinement of technologies are needed to probe and become cheaper and use fast-turnaround individual device components to integration engineer intracellular phase behavior, and standardized protocols. Data needs to of multiple signaling capabilities, to address should interface with deep proteomics, become findable, accessible, interoperable the richness of communication within metabolomics and genomics readouts of and re-usable for large-scale analyses. neural circuits. Such a paradigm shift biological function. These technologies Metabolome atlases of compound levels highlights the need for fluid exchange will also elucidate non-equilibrium driving in organs and cells are needed to compare of ideas between the fields and demands forces within the complex intracellular individual studies against animal models and understanding of fundamental physical milieu, and provide the foundation for a human population health data. Eventually, principles at the core of each technology. rigorous understanding of living matter. the community should tackle the biggest bottleneck: interpreting metabolomics data Edward Boyden: Ibrahim I. Cissé: sets by extending database queries towards Over the last few To detect a single automatic literature text mining. decades, we have fluorescent molecule, seen the invention it must either be Petra Fromme: of new technologies dilute or one must Biological for imaging brain turn off any other processes are highly activity, controlling nearby fluorescent dynamic, but most brain activity, molecule. Although biomolecular and mapping the ability to structure Credit: Matt Staley, the molecular localize individual determination Credit: Justin Knight HHMI/Janelia composition and fluorophores is approaches only wiring of the brain. advantageous and show a static picture. An important methodological challenge has led to development of super-resolution X-ray-free electron Credit: Mary Zhu will be to optimize these technologies and fluorescence microscopy, an implication of lasers (XFELs) incorporate them into a single workflow, so needing sparse fluorescent molecules is the have revolutionized that scientists can systematically investigate concentration limit of a few nano-molar or structural biology with femtosecond how the molecular composition and wiring less that it imposes. Practically, this means pulses: structures can be determined of the brain yields its emergent dynamics, that, at molecular resolution, live-cell before destruction takes place, enabling the which in turn generates behavior and fluorescent microscopes only capture discovery of the dynamics of biomolecular pathology. For example, experimental the more strongly interacting biomolecules, reactions ‘on the fly’. However, access to workflows that enable imaging activity and are blind to most assemblies of XFELs is limited, with only five facilities throughout a brain circuit, then perturbing weaker affinities. However, the growing in the world. Compact XFELs, which its dynamics, and finally mapping the appreciation for biomolecular condensates aim to shrink XFELs from 1 mile to 30 molecules and wiring throughout, may and in vivo phase transitions will likely feet, could bring XFEL technology to the yield new insights into the mechanisms force us to come up with clever ways to laboratory scale, opening the field to the underlying complex brain functions and unveil the blind spots of in vivo single- broad scientific community. Combined with dysfunctions. molecule microscopy. ultrafast spectroscopy, this will enable the

Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods 945 feature determination of the dynamics of molecular or duplicate the interdisciplinary training of our next and electronic structural transitions have been generation of life scientists will also simultaneously, in real time, in the future. characterized at the be essential. atomic level, and Anne-Claude the CRIPSR–Cas9 Grant Jay Jensen: Gingras: Proteomics genetic engineering The history of cell research is currently revolution has biology has been undergoing a burst helped dissect their punctuated by of exciting technical functions. These major advances in developments, in Credit: Institut Curie technologies provide imaging technology. terms of improving profound insights Cryo-EM imaging throughput, into how genome methods have quantification and structure relates to genome regulation and recently enjoyed an the ability to analyze gene expression. The main technological amazing ‘resolution Credit: Annie Tong, very small samples by challenge today is to follow the dynamics revolution’. In the future, the range of Sinai Health System mass spectrometry. of genome folding and function over time samples that can be imaged will expand to These improvements in living cells, integrating imaging and both much smaller and much larger targets. are already being applied to profile protein genomic data. We also have to address the For imaging macromolecules, electrons abundance, but they can also be employed behavior and role of the repetitive portion have profound advantages over X-rays in in functional proteomics. Coupled with, for of the genome, which may dictate many that they can be focused to high resolution, example, CRISPR technologies and advances of its architectural and regulatory features. revealing phases as well as amplitudes. in protein labeling and crosslinking The repeat fraction of the genome has been Because of this, and because imaging in 3D techniques, advanced proteomics methods somewhat of a blind spot for the analysis of is better than 2D, the way of the future will will provide fine details of cellular genome architecture, yet it may contain key be to image macromolecules using cryo- organization, as well as of changes in the architectural and regulatory features. electron tomography. Eventually, this will be association, localization and functions of true across scales and context from crystals proteins following perturbations. This will Stefan W. Hell: Now of small purified proteins to enormous require the acquisition of multi-faceted that the ultimate macromolecular complexes inside tissues, datasets, and one of the next challenges resolution limit but there are formidable technical challenges will be to develop tools to facilitate their in fluorescence to be overcome in sample preparation, visualization and re-use by the broader microscopy—that instrumentation and analysis. scientific community. is, 3D resolution of the size scale of Rachel Karchin: Casey S. Greene: We a molecule—has Cancer researchers are generating data been reached with are working with at an unprecedented MINFLUX, we high-dimensional scale and at levels of should seriously think beyond fluorescence. data: genomic, resolution ranging Coming up with molecular signals that are transcriptomic, from environmental as specific as fluorescence but do not require proteomic and sensors to molecular labeling with reporter molecules; that would epigenomic, from profiling of be something. bulk sequencing individual cells. It can to single-cell be tempting to search Elizabeth sequencing of tens of thousands of cells. Credit: Anna Greene through large-scale Hillman: The latest New imaging technologies will provide 2D datasets to identify microscopes are and 3D views of the cancer cells and their results that support existing notions. A near- revealing the inner environments. Longitudinal studies will term challenge is to develop techniques that workings of living make it possible to model the dynamics integrate data to illuminate under-studied organisms like never of these changes in many dimensions. We processes or reveal relationships that are at before: dynamics imagine it will be feasible to associate the odds with our expectations. Uniting machine of motion, high- dynamics of omics measurements and learning methods with representations of speed signaling, imaging with clinical outcomes for a large biomedical knowledge that account for the connectivity, and population of patients, when machine- Credit: Eileen Barroso complexity of living systems will be critical molecular and readable electronic medical records are to designing computational techniques that genetic identity, all in adopted on a large scale. To support clinical can overturn our existing understanding the context of function. Seeing is believing, decision making, we will need algorithms and sidestep confirmation bias in this era of and what used to be inferred can now be that can handle high-dimensional data and abundant data. directly observed. However, better ways that provide interpretable results. to extract quantitative information from Edith Heard: Thanks to revolutionary these datasets are urgently needed. Brilliant Laura L. Kiessling: The surface of every chromosome conformation capture and biologists with novel specimens need both cell is coated with glycans (glycoproteins, imaging technologies, we are attaining an new expertise and accessible analysis tools glycolipids and polysaccharides) that serve unprecedented understanding of genome to move beyond beautiful visualizations to as the ‘face’ of the cell, reflecting its identity architecture. The structures of many of find patterns, trends and answers. Artificial and state. In humans, glycans are critical for the protein complexes that sculpt, read intelligence will surely help, but deeper distinguishing foreign from self (for example,

946 Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods feature

microbial versus especially in perturbation experiments, techniques and the human cells) and will greatly accelerate our understanding gentlest of pipetting, diseased from healthy of which microorganisms produced which have permitted cells. Still, we cannot molecules. Third, improved tools for spatial ultra-long reads yet determine a cell’s mapping will enable visual analytics and (over 100 kb and up glycome. We must deep learning of microorganism–molecule to 2.3 megabases) develop technologies interactions, and improve our understanding to be generated that sensitively and of how microorganisms and their products from cell lines. This accurately identify exchange between hosts and environments. technique recently and sequence glycans. permitted the first New methods to elucidate the relationship Philipp Kukura: telomere-to-telomere assembly of a human between genomic data and cell-surface Single-molecule chromosome. The next big challenge is to glycans could transform our understanding of methods have make this approach applicable to human human health and disease. Such tools would had a significant clinical samples containing much smaller also illuminate the basis for cell interactions impact on the life amounts of DNA, and to find creative in tissues, host–microorganism interactions sciences, ranging bioinformatics approaches that rapidly and mixed biological communities. from imaging generate robust de novo genome assemblies and structure and enable clinical interpretation both for Benjamin P. determination to human and the microbiome! We Kleinstiver: The DNA sequencing. A are trying to tackle these problems as part unrelenting growth of central challenge for the field is applicability: of a global collaboration, so please join our the ‘CRISPR toolbox’ transforming techniques used by specialists Long Read Club. has fundamentally answering specific questions into those that altered the type are universally usable. There is something Loren Looger: and scale of genuinely unique about being able to watch Much progress biological questions single molecules come together in space has been made that the research and time: the resulting images and movies in the activation and therapeutic directly reveal the mechanisms we draw or silencing of communities can ask. Our ability to edit when we try to conceptualise complex genetically defined DNA sequences in virtually any organism biomolecular processes. Key will be to populations of cells has given humanity the technologies connect the universality of our diagrams with light, drugs, heat necessary to study life and potentially cure with the applicability of our technologies to and sound. Methods disease. We eagerly await answers from the enable the next generation of breakthroughs Credit: Matthew for the control of first CRISPR-containing human clinical in the life sciences. Staley, Janelia specific proteins trials that utilize genome editing to augment Research Campus lag far behind. immune-oncology and to treat inherited Madeline A. Ideally, techniques genetic diseases of the eye, blood, muscle Lancaster: The would: be at the and liver. Pending results that may motivate human brain, one protein, not nucleic acid, level; be essentially further tweaks and improvements to the of the final frontiers instantaneous and easily reversible; function technologies, the community may not need of exploration, still on endogenous, not over-expressed, protein; to ask what we cannot do with CRISPR remains largely a not disrupt function in the unstimulated for much longer, but instead might more mystery. How does state; and work in living animals and plants. seriously contemplate what we should such an otherwise For instance, the instantaneous, reversible not do. indiscriminate ablation of a single transcription factor or lump of protoplasm receptor in genetically defined cells (or Credit: MRC Rob Knight: The carry out advanced sub-cellular compartments) would reveal Laboratory of metagenomics human cognition? its contributions to cellular function and community is poised Brain organoids animal behavior in unprecedented detail. to make three major (models of the advances. First, developing human brain) are now allowing Emma Lundberg: an accumulation us to embark on a new age of discovery in Measuring the of reference neuroscience. The next 5–10 years will see expression of genomes (especially a rapid succession of human neurological biomolecules in space metagenome- conditions modelled with this highly and time at the single assembled genomes) relevant and tractable system. In the long cell level will deepen Credit: David Ahntholz will make reference- run, advancements in vascularisation our understanding mapping increasingly and functional connectivity will push this of cell identity. Such feasible for a wide range of environments, technology further, and have the potential studies of RNAs, allowing easier estimation of which to answer an age-old question: what makes Credit: Markus proteins, lipids genomes are in each environment, and at us human? Marcetic and metabolites what abundance, from cheap, short-read are becoming data. Second, an integration of genomic Nicholas Loman: Recent advances in increasingly feasible with chemical data (for example, short- nanopore sequencing, combined with with advanced imaging, sequencing and chain fatty acids and other metabolites), re-discovery of classical DNA extraction mass spectrometry platforms. Exciting

Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods 947 feature methodological challenges include the visualization process will be regarded a variety of development of computational models of as a reaction towards objects, and our molecular events cells that integrate molecular and spatial research efforts will lead us closer to real can profoundly alter information, and can represent cells as understanding. It is time to evaluate the protein function the dynamic and complex systems they assets of bio-imaging for their potential without affecting are. Such single cell omics methods and and limitations to truly benefit from this protein levels. A computational cellular models have the relatively new technology. key challenge for potential to revolutionize our understanding the future will be of the normal states of human cells and Eugene W. Myers to find ways to trajectories into disease. By tuning the Jr: In genome Credit: Kaska Nowak simultaneously models to represent any cellular state, sequencing, monitor all these we should be able to infer the concerted improvements in events and thus changes that allow cells to perform their technology and provide a comprehensive picture of protein functions. computer algorithms states. Protein structures integrate molecular will soon allow cues such as chemical modification, Qingming Luo: us to perfectly conformational change, interaction with Our knowledge of sequence a complex, other molecules and cleavage, which all neuroscience is based multi-gigabase affect protein function. I propose that on comprehensive genome de novo at a modest price point, detecting protein structural changes on identification and US$1,000 or less. This will herald an a global scale by mass spectrometry will characterization of unprecedented exploration of ecosystems provide novel ways to comprehensively distinct neurons and and the evolution of life. Many technical detect protein functional changes, capture neuronal circuits. and methodological challenges must first physiological and pathological alterations, Obtaining brain- be solved. In microscopy, microscopes are and generate mechanistic hypotheses. wide mammalian becoming increasingly programmable, and brain atlases at single-neuron resolution ‘smart’ devices and computational methods Wolf Reik: We with identified neuron morphology and such as deep neural nets are enabling us to are witnessing an entire neuronal circuits containing long see further and more clearly into biological enormous revolution projections is still challenging and requires samples. A key challenge is to fully harness in single-cell the development of wide-field imaging the power of adaptive optics, particularly genomics, which techniques with high throughput and in devices and samples where the use of is being applied to high voxel resolution, as well as intelligent fiducial markers and explicit measurement millions of cells high-throughput mass data processing of the wave front are not possible. and giving rise to techniques. Once we retrieve the entire set a new anatomy of of projections of specific neuronal circuits Garry P. Nolan: the human body as well as the affiliated functionally defined Single-cell through the Human Cell Atlas and Human brain areas (which we call brainsmatics), it phenotyping is Developmental Biology initiatives. But will be exciting to unravel mysteries such as moving towards there are many more layers of molecular the mechanisms of consciousness, dreams generating tissue information we can capture now and in and cognition. Those discoveries will benefit atlases and trekking the future in single cells, combining the our understanding of and development of inward towards transcriptome with DNA modifications therapy strategies for neurological disorders. establishing a 3D and chromatin accessibility, histone map of a cell’s marks and perhaps the proteome as well. Atsushi Miyawaki: constituents, An integration of time as a dimension in The introduction simultaneously driving algorithmic these measurements would be particularly of functional development that enables mere humans exciting. Powerful machine-learning probes may lead to to understand biology. The limitations of algorithms will connect these layers either the up- or current marker technologies, including together and will be able to detect cell fate down-regulation antibodies, chemical tags and gene fusions, decisions, or cell fate change in disease, at an of downstream beg the question of how do we measure unprecedented level of precision. Eventually, intracellular everything? Inevitably, we need every single-cell editing may allow pathological signaling, and atom’s position and identity, and from that changes in cell fate to be corrected, although Credit: RIKEN CBS may perturb the atom cloud reconstruct the identities and this may take a little while yet. cells we observe. positions of all cellular constituents. We Moreover, even with are developing the concepts behind such Markus Sauer: knock-in methods for probe introduction, an instrument to determine the positions Super-resolution a substantial amount of light or chemicals of every atom in situ at sub-Ångstrom microscopy methods are absorbed by cells labeled with resolution. The field has spent so much can provide spatial fluorescent or bioluminescent probes, time inferring, indirectly, a cell’s interior resolution that is well respectively. Quantitative bio-imaging structure; why not just take a picture? below the diffraction- is expected to provide a methodological limit of light framework for simulating observation- Paola Picotti: Proteomics can measure microscopy, but they dependent perturbation. Once we accept changes in the abundances of proteins for do not yet provide the the idea that ‘seeing is perturbing’, the almost complete proteomes. However, molecular resolution

948 Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods feature required to understand how a cell functions genomic screens in vivo. Encouraging proof- for many genetic and which mechanisms occur in the case of-concepts have recently been described for variants. In the of a dysfunction or disease. I anticipate at least some of these goals. next decade, the that within the next years, combinations challenge will be to of methods such as expansion and super- Nikolai Slavov: integrate regulatory resolution microscopy, supported by the Recently, mass- and coding variant development of intelligent dyes and labeling spectrometry effects across the methods with minimal linkage error, will methods have whole genome provide imaging of organelles and protein increased the to holistically complexes with one to two nanometer specificity and Credit: Ruth predict phenotypic resolution. By harnessing these tools, the throughput of Dannenfelser consequences future will allow us to decipher how nature quantifying proteins for patients. This encodes function at the molecular level. in single mammalian requires advances in cells: we can now modeling approaches as well as improved Credit: Ivana Dimitrova Alex K. Shalek: quantify thousands algorithmic efficiency, scalability and model Single-cell RNA-seq of proteins across interpretation. Critically, all progress relies has transformed hundreds of single cells. I am confident on continued generation and sharing of our ability to that soon we will extend these methods to experimental and clinical data. Integrative dissect cellular quantifying metabolites, post-translational whole genome interpretation will deepen systems, enabling modifications, and the dynamics and our understanding of genetics and can transcriptome-wide spatial distributions of proteins and their transform our ability to precisely diagnose identification of complexes. Ultimately, the accuracy, and treat diverse diseases. cellular components completeness and throughput of these and their molecular measurements will provide data for David van Valen: Credit: Juliana Sohn signatures. Yet, transitioning from descriptive classification The intersection we still need to do of single cells to quantitative models of of deep learning more, such as: faithfully capture cell states regulatory protein interactions. I believe and biology is a at scale to decipher critical molecular these data and models will enable systematic very exciting space, attributes; comprehensively appreciate what inference of direct causal mechanisms that particularly for those a ‘transcriptional snapshot’ can actually tell underpin biological functions. of us who work with us about a cell’s past, present and future biological images, within a tissue; and systematically uncover Amos Tanay: as these methods the value derived from collecting and Epigenomics is are starting to integrating additional data (for example, moving toward provide robust solutions to long-standing spatial position, dynamics, other omics, single-cell problems such as image restoration, image existing single-cell datasets, reference gene resolution and is segmentation and object tracking. To signatures and perturbations). Equally already facilitating me, the most exciting aspect of this area important, we must also empower global unprecedentedly is seeing the creative ways biologists are participation in the generation and analysis sharp descriptive incorporating deep learning throughout of these data to achieve broad mechanistic analysis of multiple their experimental designs and analytics insights into human health and disease. epigenetic scales, pipelines. As deep learning methods become ranging from DNA methylation, through more commonplace, I think we are going Jay Shendure: This is decorated nucleosomes, up to chromosomal to see a drastic increase in the pace of a very exciting time topologies. But regulates genes biological discovery. for high-throughput by changing their physical contexts rather functional genomic than turning them on and off in a digital Hong-Wei Wang: screens. The fashion. Understanding all its scales and Cryo-EM uses growing CRISPR layers, therefore, requires truly quantitative transmission toolset is enabling models that are based on millions of single- electron microscopy increasingly versatile cell epigenomic profiles. Such models must to study frozen- experiments, for go significantly beyond black-box machine- hydrated specimens example, expanding learning predictions. We will have to learn to at liquid nitrogen the ‘targetable genome’ to noncoding use the new data to develop principled and temperature to regions. In my view, the primary challenge interpretable tools, with a clear multi-scale reveal the structures of the moment lies with expanding the biophysical basis that can match the of macromolecules range of phenotypes that are compatible multi-scale biology of the genome and or cellular organelles in their relatively with such screens beyond the typical its regulation. close-to-native states. The recent ‘growth rate’ experiments. This includes, hardware and software breakthroughs in but is not limited to, whole transcriptional Olga Troyanskaya: With the broad cryo-EM technology have transformed or epigenetic profiling, as well as imaging- availability of whole genome sequencing, structural biology to a new phase, based phenotyping, in association with the promise of precision medicine relies on where macromolecule structure can each perturbation. Further challenges the comprehensive interpretation of these be more robustly elucidated at near- include achieving comprehensive pairwise genomes. Recently, deep learning models atomic resolution. Instrumentation and interaction screens and moving functional enabled the prediction of regulatory effects computational developments

Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods 949 feature of cryo-EM methods in the next Magdalena Olga Troyanskaya63,64, David van Valen65, decade will aim to solve structures at Zernicka-Goetz: Hong-Wei Wang66, Chengqi Yi67, resolutions close to 1 Ångstrom, deciphering One of the most Peng Yin68,69, Magdalena Zernicka-Goetz70,71 the dynamic conformational landscapes interesting challenges and Xiaowei Zhuang72 of macromolecules during reactions, in my field would 1Departments of Materials Science & Engineering and revealing high-resolution molecular be to uncover the Brain & Cognitive Sciences, Massachusetts Institute of structures in situ, and directly correlating principles by which Technology, Cambridge, MA, USA. 2Research structures with functions in a broader the embryo builds Laboratory of Electronics, Massachusetts Institute of cellular context. itself so that we Technology, Cambridge, MA, USA. 3McGovern Institute for Brain Research, Massachusetts Institute Credit: David Glover can create embryo Chengqi Yi: models from of Technology, Cambridge, MA, USA. 4Department of Epitranscriptomic cultured stem cells. Neurotechnology, Massachusetts Institute of sequencing Such models, if successful, would provide Technology, Cambridge, MA, USA. 5MIT Media Lab, technologies powerful tools to understand the complexity Massachusetts Institute of Technology, Cambridge, that enable of intrinsic interactions between the cells MA, USA. 6Department of Chemical and Biological transcriptome-wide that are essential for the embryo-building Engineering, Princeton University and Howard mapping of RNA process, with its distinct organs, as well as Hughes Medical Institute, Princeton, NJ, USA. modifications have uncover how developmental defects arise 7Department of Physics, Massachusetts Institute of added valuable and how we can prevent them. Of course, Technology, Cambridge, MA, USA. 8West Coast knowledge about such research has to be bounded and guided Metabolomics Center, University of California Davis, the role and regulation of RNA. Yet, by ethical considerations. Davis, CA, USA. 9Biodesign Center for Applied there is an unmet biological need to Structural Discovery and School of Molecular quantify the absolute stoichiometry of Xiaowei Zhuang: Sciences, Arizona State University, Tempe, AZ, USA. the epitranscriptome. In addition, robust With recent 10Lunenfeld-Tanenbaum Research Institute, Sinai and sensitive methods that are highly advances in imaging Health System, Toronto, Ontario, Canada. reproducible and can serve as the gold and genomics 11Department of Systems Pharmacology and standard of detection are still lacking for technologies, it Translational Terapeutics, Perelman School of the majority of RNA modifications. Tools is truly exciting Medicine, University of Pennsylvania, Philadelphia, to specifically manipulate epitranscriptomic to envision the PA, USA. 12Childhood Cancer Data Lab, Alex’s marks in spatially and temporally controlled possibility of two Lemonade Stand Foundation, Philadelphia, PA, USA. manners are also urgently needed. Future previously seemingly 13European Molecular Biology Laboratory, Heidelberg, challenges and exciting opportunities unreachable goals. Germany. 14Collège de France, Paris, France. 15Max include epitranscriptome analysis at the The first is to generate a full census and Planck Institute for Biophysical Chemistry, Göttingen, single-cell and single-molecule level, and atlas of cells for living organisms, including Germany. 16Max Planck Institute for Medical in situ via the combination of sequencing human beings. Although the scale may Research, Heidelberg, Germany. 17Departments of and imaging. seem daunting — a human is made of tens Biomedical Engineering and Radiology, Columbia of trillions of cells — the rapid development University, New York, NY, USA. 18Mortimer B. Peng Yin: DNA of single-cell omics methods, including Zuckerman Mind Brain Behavior Institute, Columbia nanotechnology image-based single-cell transcriptomics, University, New York, NY, USA. 19Departments of enables precise will allow this goal to be achieved in the Biology and Biophysics, California Institute of engineering of foreseeable future. The second is to generate Technology and Howard Hughes Medical Institute, nanostrucures with a full molecular architecture of the cell. The Pasadena, CA, USA. 20Department of Biomedical user-prescribed advent of super-resolution imaging and Engineering, Te Johns Hopkins University, structural and genomic-scale imaging has led us closer Baltimore, Maryland, USA. 21Department of dynamic properties, to realizing this ambition, though major Oncology, Johns Hopkins Medical Institutions, and has recently challenges still lie ahead, making this a Baltimore, Maryland, USA. 22Te Institute for advanced diverse longer-term goal. ❐ Computational Medicine, Te Johns Hopkins Credit: Seth Kroll bioimaging University, Baltimore, Maryland, USA. 23Department approaches by Polina Anikeeva1,2,3, Edward Boyden3,4,5, of Computer Science, Te Johns Hopkins University, providing enhanced Cliford Brangwynne6, Ibrahim I. Cissé7, Baltimore, Maryland, USA. 24Department of resolution, signal amplification and Oliver Fiehn8, Petra Fromme9, Chemistry, Massachusetts Institute of Technology, multiplexing abilities, as well as methods in Anne-Claude Gingras10, Casey S. Greene11,12, Cambridge, MA, USA. 25Center for Genomic biosensing and single-molecule biophysics. Edith Heard13,14, Stefan W. Hell15,16, Medicine, Massachusetts General Hospital, Boston, More sophisticated nanodevices that Elizabeth Hillman17,18, Grant Jay Jensen19, MA, USA. 26Department of Pathology, Massachusetts perform in situ analysis of the molecular Rachel Karchin20,21,22,23, Laura L. Kiessling24, General Hospital, Boston, MA, USA. 27Department of environment to generate real-time signal Benjamin P. Kleinstiver25,26,27, Pathology, Harvard Medical School, Boston, MA, or action, or to encode spatial temporal Rob Knight28,29,30,31, Philipp Kukura32, USA. 28Department of Pediatrics, University of features in DNA records, are particularly Madeline A. Lancaster33, Nicholas Loman34, California, San Diego, La Jolla, CA, USA. exciting for future development. Dare Loren Looger35, Emma Lundberg36,37,38, 29Department of Bioengineering, University of we even imagine molecular robots that Qingming Luo39,40, Atsushi Miyawaki41,42, California, San Diego, La Jolla, CA, USA. survey an otherwise inaccessible molecular Eugene W. Myers Jr.43,44,45, Garry P. Nolan46, 30Department of Computer Science & Engineering, landscape, in a similar spirit as Web crawlers Paola Picotti47, Wolf Reik48,49,50, Markus Sauer51, University of California, San Diego, La Jolla, CA, that index the internet or Mars rovers that Alex K. Shalek24,52,53,54,55,56,57, Jay Shendure58,59, USA. 31Center for Microbiome Innovation, University inspect the planetary surface? Nikolai Slavov60,61, Amos Tanay62, of California, San Diego, La Jolla, CA, USA.

950 Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods feature

32Physical and Teoretical Chemistry Laboratory, of Microbiology & Immunology, Stanford University 62Departments of Computer Science & Applied Department of Chemistry, University of Oxford, School of Medicine, Stanford, CA, USA. 47Institute of Mathematics and Biological Regulation, Weizmann Oxford, UK. 33MRC Laboratory of Molecular Biology, Molecular Systems Biology, Department of Biology, Institute of Science, Rehovot, Israel. 63Department of Cambridge, UK. 34Institute of Microbiology and ETH Zurich, Zurich, Switzerland. 48Babraham Computer Science, Lewis-Sigler Institute for Infection, University of Birmingham, Birmingham, Institute, Babraham, UK. 49Sanger Institute, Hinxton, Integrative Genomics, Princeton University, UK. 35Janelia Research Campus, Howard Hughes UK. 50University of Cambridge, Cambridge, UK. Princeton, NJ, USA. 64Department of Genomics, Medical Institute, Ashburn, VA, USA. 36Science for 51Department of Biotechnology and Biophysics, Flatiron Institute, Simons Foundation, New York City, Life Laboratory, School of Engineering Sciences in Biocenter, University of Würzburg, Würzburg, NY, USA. 65Division of Biology and Biological Chemistry, Biotechnology and Health, KTH Royal Germany. 52Institute for Medical Engineering and Engineering, California Institute of Technology, Institute of Technology, Stockholm, Sweden. Science, Massachusetts Institute of Technology, Pasadena, CA, USA. 66School of Life Sciences, 37Department of Genetics, Stanford University, Cambridge, MA, USA. 53Koch Institute for Integrative Tsinghua University, Beijing, China. 67Peking Stanford, CA, USA. 38Chan Zuckerberg Biohub, San Cancer Research, Massachusetts Institute of University, Beijing, China. 68Wyss Institute for Francisco, CA, USA. 39School of Biomedical Technology, Cambridge, MA, USA. 54Broad Institute Biologically Inspired Engineering, Harvard University, Engineering, Hainan University, Haikou, China. of MIT and Harvard, Cambridge, MA, USA. 55Ragon Boston, MA, USA. 69Department of Systems Biology, 40Wuhan National Laboratory for Optoelectronics, Institute of MGH, MIT and Harvard, Cambridge, Harvard Medical School, Boston, MA, USA. Huazhong University of Science and Technology, MA, USA. 56Division of Health Sciences and 70Division of Biology, California Institute of Wuhan, China. 41Laboratory for Cell Function Technology, Department of Immunology, Harvard Technology, Pasadena, CA, USA. 71Department of Dynamics, Brain Science Institute, RIKEN, Wako, Medical School, Boston, MA, USA. 57Department of Physiology, Development and Neuroscience, Japan. 42Biotechnological Optics Research Team, Immunology, Massachusetts General Hospital, , Cambridge, UK. Center for Advanced Photonics, RIKEN, Wako, Japan. Boston, MA, USA. 58Genome Sciences, University of 72Departments of Chemistry & Chemical Biology and 43Center for Systems Biology Dresden, Dresden, Washington, Seattle, WA, USA. 59Brotman Baty Physics, Harvard University and Howard Hughes Germany. 44Max Planck Institute of Molecular Cell Institute for Precision Medicine, Seattle, WA, USA. Medical Institute, Cambridge, MA, USA. Biology and Genetics, Dresden, Germany. 60Department of Bioengineering, Northeastern 45Department of Computer Science, Technical University, Boston, MA, USA. 61Barnett Institute, Published online: 27 September 2019 University Dresden, Dresden, Germany. 46Department Northeastern University, Boston, MA, USA. https://doi.org/10.1038/s41592-019-0585-6

Changing the way you see life Ultra Precise Motion Control - D.C. Servo motors down to 20 nm, piezos down to 1 nm, and low drift XYZ stages.

Microscopy - Automation, modular microscopes, autofocus complete light sheet systems, and components. OEM - Custom designed systems to APPLIED SCIENTIFIC user specifications. INSTRUMENTATION

www.asiimaging.com • [email protected] (800) 706-2284 or (541) 461-8181

Nature Methods | VOL 16 | OCTOBER 2019 | 945–951 | www.nature.com/naturemethods 951