The Somatic Mosaicism across Human Tissues Network

Abstract

From fertilization onwards, the cells of the human body acquire variations in their DNA sequence, known as somatic mutations. These post-zygotic mutations arise from intrinsic errors in DNA replication and repair, as well as from exposure to mutagens. Somatic mutations have been implicated in some diseases, but a fundamental understanding of the frequency, type and patterns of mutations across healthy human tissues has been limited. This is primarily due to the small proportion of cells harboring specific somatic variants within an individual, making them more challenging to detect than inherited variants. Here, we describe the Somatic Mosaicism across Human Tissues (SMaHT) Network, which aims to create a reference catalog of somatic mutations and their clonal patterns across 19 different tissue sites from 150 non-diseased donors and develop new technologies and computational tools to detect somatic mutations and assess their phenotypic consequences, including clonal expansions. This strategy enables a comprehensive examination of the mutational landscape across the human body, and provides a comparison baseline for somatic mutation in diseases. This will lead to a deep understanding of somatic mutations and clonal expansions across the lifespan, as well as their roles in health, aging, and, by comparison, in diseases.

Introduction

Genetic diversity within the human population has been well described. The Human Genome Project resulted in the first near-complete mapping of the human DNA sequence¹, and was followed by large-scale projects, such as the 1,000 Genomes Project² and the Pangenome project³, that mapped the genetic diversity between individuals and populations. Now, there is growing recognition that extensive genetic variation exists within individuals among different tissues and cells. Two decades after completion of the first draft human genome, the Somatic Mosaicism across Human Tissues (SMaHT) Network plans to map the genetic diversity across different tissues and cells within individuals.

From fertilization onwards, the cells of the human body continuously experience damage to their genome, either from intrinsic causes or from exposure to mutagens^4–9. While the vast majority of DNA damage is repaired, and the genome is replicated with extremely high fidelity, cells steadily acquire somatic mutations throughout life. All cells within an individual harbor somatic mutation, but any given mutation is present in only a subset of the cells, or even in single cells. Hence, somatic mutations are often described as mosaic^10,11.

The detection of somatic mutations is challenging. In contrast to inherited variants, somatic mutations only exist in small and variable proportions of cells, ranging from embryonic mutations present in most cells down to mutations present in just a single cell (Fig. 1a). This challenge is exacerbated by the introduction of artifacts and errors resembling low-frequency mutations during DNA library preparation and sequencing¹². Current short-read sequencing technologies limit detection of mutations in repetitive regions of the genome and are likely to be less suitable for detection of somatic structural variations.

Figure 1 |. Somatic mutations, causes and patterns:

a, Schematic comparison between inherited variants, an early somatic mutation and a late somatic mutation. b, Overview of causes and types of somatic mutations. RT=reverse transcriptase, EN=endonuclease c, Overview of the reported mutation rates of somatic SNV across developmental stages and tissues. Data of first cell divisions^3,4,6,40 and later cell divisions^3,4,6,40 are SNVs per cell per division; ZGA, zygotic genome activation. Data from fetal development of the early central nervous system (CNS)⁶ and placenta⁴¹ are SNVs per cell per day. Adult data is SNVs per year and estimated for seminiferous tubules⁴², hematopoietic stem cells^26,43,44, B lymphocytes⁴³, neurons^21,30,45, T lymphocytes⁴³, bronchial epithelium⁴⁶, gastric epithelium⁴⁷, endometrial epithelium⁴⁸, hepatocytes¹⁹, small bowel epithelium^19,49, colorectal epithelium^19,24,29, cardiomyocytes⁵⁰.

While most somatic mutations are likely functionally neutral¹³, some can profoundly alter the phenotype of a cell and are implicated in a wide variety of diseases. Many insights have come from sequencing the genomes of cancers¹⁴, the best-known example of disease arising from somatic mutation, but mutagenesis in tumors is often accelerated, and normal mutational patterns are distorted by genome instability. More recently, mapping the patterns of somatic mutations in normal tissues, exemplified by efforts of the Brain Somatic Mosaicism Network (BSMN), and other studies^{3–6,15–26}, has identified a role for somatic mutations in developmental syndromes, neurological diseases, and inflammatory disorders^27–39. Despite these efforts, there is currently no robust reference data set of somatic mosaicism across many tissues of a large pool of donors.

In this Perspective, we describe the SMaHT Network, initiated by the NIH Common Fund, which aims to generate a reference catalog of somatic variation from 150 donors in 19 non-diseased tissue sites. To advance the field beyond what is currently known, the SMaHT Network will perform a comprehensive discovery and analysis of all types of somatic mutations at an unprecedented scale: the joint analysis of mosaicism across many tissues per donor; the robust discovery of structural variants through long-read sequencing and donor-specific assemblies; the widespread and robust application of ultrasensitive sequencing technologies such as Duplex sequencing across sequencing centers. Furthermore, beyond applying established sequencing assays at scale, the SMaHT Network has a strong emphasis on tool and technology development to enable the next generation of somatic mutation studies. Before describing the network goals in detail, we briefly review the current knowledge about somatic mutations in health and disease, as well as the technical challenges in mutation detection. A large part of the SMaHT Network will focus on the development of technologies and computational tools to improve the detection of all types of somatic variation.

Somatic mutations in healthy tissues

Throughout the human lifespan, from conception to death, cells acquire mutations in their DNA (Fig. 1b)^1–4,6,51. These somatic mutations can be the consequence of erroneous repair of damaged DNA bases or DNA strand breaks, errors during replication, chromosome missegregation, or the integration of mobile elements. Somatic mutations can be divided into different types⁵²: substitutions, the vast majority of which are single nucleotide variants (SNVs); small (<50 base pairs) insertions and deletions (indels); structural variants (SVs), including segmental duplications, large deletions, translocations, inversions, mobile element insertions (MEIs) and complex SVs, including chromothripsis and chromoplexy; and other large chromosomal aberrations, such as whole-chromosome gains and losses. Duplications, deletions and whole-chromosome gains and losses are also referred to as copy number variants (CNVs) or mosaic chromosomal alterations. These classes differ profoundly in their underlying causes and patterns across tissues and their phenotypic effects on cells. In normal tissues, SNVs are by far the most common type of somatic variation, followed by indels. SVs and large chromosomal aberrations are observed less frequently²⁷, but typically affect more base pairs and thus may have larger functional effects. However, most previous studies on somatic mutations relied on short-read DNA sequencing, which may fail to detect various types of SVs. Studies of germline differences has shown SVs are far more abundant, but the majority are missed by short-read approaches⁵³.

Different mutagenic processes cause distinct patterns of somatic mutations, depending on the types of DNA damage incurred and the pathways responsible for DNA lesion repair. Research over the past decade has deconvolved these patterns into mutational signatures and linked certain signatures to specific mutagens, such as ultraviolet light, tobacco smoke, chemotherapy, or natural age-related accumulation of endogenous mutations⁵⁴. Mutational signatures are most commonly applied to SNVs⁵⁵ but they have been defined for other classes of somatic mutations, including indels⁵⁵, chromosomal alterations^56,57 and SVs^55,58. In the context of SNVs, mutational signatures reflect the distribution of specific base changes within their trinucleotide contexts.

All normal tissues, including post-mitotic cells, exhibit SNV mutational signatures linked to clock-like endogenous processes (single base substitution signature 1 (SBS1) or SBS5) and, to a lesser extent, oxidative damage (SBS18)^42,50,59. Mutational signatures linked to mutagenic exposure can be confined to specific organs, such as UV damage (SBS7) in the skin⁶⁰ or skin-resident T-lymphocytes⁴³, damage from tobacco smoke (SBS4) in the bronchial epithelium of the lung⁴⁶, and exposure to a genotoxic strain of E. coli (SBS88)²⁴ in the large intestine. These exposure differences drive some of the variation in the types of somatic mutations observed across different tissues of the human body^3,42,61. Further, different mutational processes show different correlations with genomic features, such as replication timing, replication strand and transcription strand^62–65, reflecting genomic biases of DNA damage and repair.

The somatic mutation rate varies across human tissues and life stages (Fig. 1c). During the initial embryonic cell divisions, somatic SNVs accumulate at a high rate of approximately 3 per division, likely due to the high division rate and delayed activation of the zygotic genome^3,4,40. Afterwards the mutation rate decreases (~1 SNVs per division) during development in utero, both in embryonic, such as the fetal brain^6,66,67, and extraembryonic tissues, such as the placenta⁴¹. After birth, mutation rates further decline 5- to 10-fold and vary substantially across tissues, from 16-20 SNVs per year in post-mitotic cells such as neurons^21,45,59,67 to 44 SNVs per year in colonic stem cells²⁴ (Fig. 1c). Germ cells have the lowest somatic mutation rate reported²³, in line with the parental age effect on de novo germline mutations⁴². While division rate may influence the endogenous somatic mutation rate, there are likely other factors that modulate both mutagenesis and repair of DNA damage^68–70.

Large somatic mutations such as SVs, chromosomal alterations and MEIs are detected at much less frequently than SNVs and indels. While somatic aneuploidy appears to be rare, sub-chromosomal structural variations affect 13-41% of neurons^18,34,71,72. Frequent CNVs, mostly duplications, of likely developmental origin have been detected in approximately 7% of brains from the BSMN consortium³⁴, and mosaic chromosomal alterations were observed in approximately 5% of blood samples in the UK Biobank⁷³. Single-neuron DNA sequencing of mobile element-enriched libraries or whole genomes has revealed MEI events that appear to occur during development and create mosaicism in the human brain^5,74,75. Bulk sequencing approaches have also detected a few examples of somatic MEIs in the brain⁷⁶ and non-brain tissues including the heart⁷⁷, fibroblasts⁷⁷, and liver⁷⁸. Recent somatic MEI profiling in colorectal epithelial single-cell clones has indicated peak insertion rates during early embryogenesis⁷⁹. Considering the potential impact of these large mutations on the sequence, splicing, or expression of genes^80,81, it is valuable to understand their prevalence across human tissues during development and aging.

While most somatic mutations do not discernibly affect the phenotype of a cell, some somatic mutations are under selection in different tissues. Such driver mutations may lead to a proliferative advantage or increased survival of the cell and its progeny, resulting in clonal expansions in tissues. Cancer is the canonical example of somatic evolution and often involves the stepwise accumulation of key somatic mutations and genomic instability^1,82. Mutations typically associated with cancer can be abundant across normal tissues with age. For comparison, in a typical individual of age 60, approximately 90% of the endometrial epithelium harbors a driver mutation⁴⁸, whereas this is true of only about 1% of colonic epithelium²⁴, despite the latter having a much higher somatic mutation rate^24,26. This difference is likely caused by the menstrual cycles of shedding and regrowth in the endometrium. Likely due to similar clonal expansion in development or aging, about 6% of individuals harbor a 3 to 20-fold higher than average number of detectable SNVs in their brain⁶. These varying proportions of clonally expanded cell populations likely reflect differences in tissue architecture, cell turnover, regeneration, and selection pressures, but much is still unknown.

While many driver mutations in normal tissues can be identical to those found in corresponding cancer types, their abundance and phenotypic consequence may differ profoundly as normal tissues may experience different selection pressures than cancer. For example, clones with NOTCH1 mutations are exceedingly abundant in normal esophageal epithelium, at even higher rates than esophageal cancers⁸³. NOTCH1 mutant clones have a lower propensity of malignant transformation and even outcompete precancerous clones in the esophagus^84,85. These observations suggest that characterizing the somatic mutation landscape in normal individuals will be important to understand the role of these mutations in pathological phenomena such as cancer.

Lastly, somatic mutations can be used as intrinsic barcodes to create phylogenies and trace the ancestries of cells, such that it becomes possible to quantitatively study human development from somatic mutations ascertained in adult donors^{3,4,6,20,25,51,65,66,74,86}. This approach has been applied to studying embryogenesis, clonal expansions across the lifespan and the origins of childhood cancers⁸⁷. Since the allele frequency of a mutation reflects the fraction of cells within a population that harbors it, this method can be used to quantitatively assess the contribution of embryonic progenitors to the adult body. Intriguingly, such studies have found that one of the two daughter cells of the zygote often has at least twice as many descendant cells as the other^{3,4,25,66,67,88,89}, likely due to cellular bottlenecks in embryogenesis, developmental cell death, or migratory patterns, and confirming earlier observations in mice^51,90.

Taken together, these initial studies on somatic mutations in normal tissues have shown the variability of rates, patterns, and selection of mutations across tissues. It is unknown, however, how variable these patterns are between individuals and how different types of somatic mutations are correlated with inherited genetic background, environmental exposures, or other behavioral characteristics. In addition, mutation discovery is severely hampered in poorly mapped regions of the genome, including acrocentric chromosomes, centromeric and repetitive regions, and hence, the mutational patterns in these regions are largely unknown. Thus, identification of the differences in mutational patterns between tissues and individuals, particularly in the context of specific organs^{21,26,29,30,34,91,92}, may have profound clinical implications.

Somatic mutations and disease

Somatic mutations can profoundly alter the phenotype of a cell and have been implicated in human diseases. Besides cancer, various other diseases and conditions can be a result of somatic mutations, including cardiovascular anomalies, immunological and neurological disorders^{26,30–34,43,93,94}. Notably, early somatic mutations can cause clonal expansions and alterations in the differentiation programs of precursor cells that subsequently can lead to paediatric cancers and organ overgrowth^87,95,96. Among the first described instances of somatic mutagenesis, PI3K-AKT-mTOR pathway mutations involving the brain, were associated with brain malformations leading to intractable epilepsy^33,97,98. Other examples are NRAS mutations leading to congenital melanocytic nevi⁹⁹ and UBA1 mutations in hematopoietic stem cells¹⁰⁰ leading to VEXAS syndrome, a rare and severe inflammatory disorder. Somatic expansions of short tandem repeats in the brain can cause cell death and neurodegeneration¹⁰¹, and underpin Huntington’s disease^102,103. Large SVs, including CNVs and MEIs, have been also implicated in neurodevelopmental and neurodegenerative disorders^76,104,105.

The effects of somatic mutations can be highly specific to the timing and tissue of origin. For example, an activating PIK3CA mutation acquired during development can lead to widespread overgrowth across organs and vascular malformations⁹⁴. However, PIK3CA mutations acquired after development can lead to cavernomas in the brain¹⁰⁶ and are also a common driver mutation observed in normal colonic²⁴ and endometrial epithelium⁴⁸.

Clonal expansions can also indirectly lead to or influence other diseases²⁶. An example is clonal hematopoiesis of indeterminate potential (CHIP), characterized by a clonal expansion within the hematopoietic stem cell compartment driven by somatic mutations. CHIP is highly prevalent in the context of normal aging²⁶. Besides acting as a potential cancer precursor clone, CHIP has been linked to a variety of non-cancer diseases, such an increased risk of cardiovascular disease¹⁰⁷ and infections¹⁰⁸.

Conversely, diseases can also select clones with certain adaptive somatic mutations. Recently, it has been shown that inflammatory bowel disease leads to the preferential remodeling of the colonic epithelium with clones harboring IL17 and Toll-like receptor pathway mutations^29,109. Likewise, chronic liver disease selects for clones of hepatocytes that escape the toxicity imposed by the disease, notably by recurrent, independent mutations in FOXO1, CIDEB and GPAM, all involved in lipid metabolism⁹².

Taken together, research over the past years has shown that somatic evolution is ubiquitous in normal tissues and is fundamental to our understanding of the causes, mechanisms, and consequences of disease, and the normal process of aging.

The SMaHT Network

The Somatic Mosaicism across Human Tissues (SMaHT) Network, funded by the NIH Common Fund, was established with the goal of transforming our understanding of how somatic variation in human cells influences biological processes. SMaHT will accomplish this through the following aims: 1) generate a comprehensive dataset of somatic variants across human tissues (Fig. 2); 2) develop tools and technologies to optimize detection and characterization of various types of somatic variants; and 3) create a somatic mutation database that is widely used by researchers and the wider public, and interoperable with similar data sets.

Figure 2 |. Tissue sampling:

Overview of sampling from 19 primary tissue sites, spanning three developmental germ layers (endoderm, mesoderm, and ectoderm) and germ cells. While organs represent mixture of cells derived from the germ layers (e.g, skin epidermis (ectoderm) versus dermis (mesoderm) and adrenal gland medulla (ectoderm) versus cortex (mesoderm), we have indicated the major germ layer represented by each organ. Gonads represent germ cells and their supportive structures (mesoderm), while buccal swabs are variable mixtures of germ layers (mesoderm and ectoderm).

The Network is comprised of five Genome Characterization Centers (GCCs), 14 Tool and Technology Development projects (TTDs), an Organizational Center (OC), a Data Analysis Center (DAC) and a Tissue Procurement Center (TPC) and includes over 250 researchers from 52 institutions. The GCCs are tasked with producing a core dataset of somatic mutations for the SMaHT Network from multiple tissues collected by TPC, while TTDs are tasked with developing novel experimental assays and computational tools. The DAC will integrate the data generated by GCCs and TTDs to build the somatic mutation catalog, data portal, and the analysis work bench for the Network. The OC will coordinate the Network activities and focus on outreach efforts and building liaison with other genomics consortia. The SMaHT Network has implemented a set of policies (https://smaht.org/policies/), including a policy to allow external researchers to apply for associate membership of the Network.

The tissues to be profiled by the Network include those arising from the three germ layers and germlines within the human body, which will give the opportunity to delineate early somatic mutations that are common across all tissues, as well as later mutations that are unique to certain tissues (Fig. 2). The TPC is partnering with multiple organ procurement organizations (OPOs) in the US for the screening, authorization, and recovery of tissues from post-mortem organ and tissue donors. Tissues will be collected following transplant recovery and include ascending and descending colon, esophagus, lung and liver (predominantly endoderm); blood, heart, aorta and skeletal muscle (predominantly mesoderm); brain, adrenal gland, sun-exposed and non-sun-exposed skin (predominantly ectoderm). We also aim to collect buccal swabs to assess the extent of the somatic mutation landscape that can be gleaned from clinically accessible tissues in living donors. To study mutagenesis in germ cells, we also aim to collect ovaries and testes. Lastly, to enable a variety of experimental techniques requiring live cells, we will derive fibroblast cultures from dermis (skin). All tissues are requested to be recovered from each donor approached for the SMaHT tissue collection. The number and type of samples collected from each donor will vary based on donor authorization and eligibility (Box 1), but the goal is to recover as many tissues from a single donor as possible. To study the mechanisms and consequences of somatic mosaicism across the lifespan, these post-mortem donors will span the human adult age ranges from 18 to over 85. Race and ethnicity of donors are assessed using a single-question framework.

Box 1.

Inclusion criteria:

• Donor over 18 years old

• Collection can be completed within 24 hours of cross-clamp or cardiac cessation

Exclusion criteria:

• History of HIV, HCV or HBV

• History of IV drug use in last 5 years

• Chemotherapy or radiation treatment in the past 24 months

• Known chromosomal or genetic disorder

• Current positive blood cultures (sepsis)

• Active/metastatic cancer

• Diagnosed with multisystem organ failure

• Received organ or allogeneic bone marrow transplant

• Received whole blood transfusion in 48 hours prior to cross clamp or cardiac cessation

Exclusion criteria for brain donation specifically:

• Cause of death related to penetrating brain injury or head trauma

• Brain dead or ventilator dependent for greater than 24 hours

To maximize scientific and clinical impact of the data set, the TPC will collect a large amount of donor metadata during donation and biospecimen collection, building on practices developed for the Genotype-Tissue Expression (GTEx)¹¹⁰ and developmental GTEx projects¹¹¹. De-identified donor-level data will include demographic information, medical history, sample-based laboratory test results and death circumstances. Sample-level data will include tissue type and location, ischemic time, and tissue metrics from pathology review. Pathology images will be made publicly available. When possible, tissue sampling will align with the common coordinate framework structure of other large-scale projects. For all of these biospecimens, sufficient fresh-frozen material will be collected and banked to enable all core assays as well as implementation of novel emerging technologies. Fixed samples for pathology review will be collected from adjacent sites to the fresh-frozen specimens utilizing a standardized collection schema developed for each tissue type.

To pursue a demographically diverse and evenly sex-distributed pool of donors, the SMaHT Network includes an Ethical, Legal and Social Implications (ELSI) project¹¹² to promote diversity, equity, and inclusion efforts. In a recent call to action, the American Society for Human Genetics stated that “addressing underrepresentation in human genomics starts with meaningful engagement of underrepresented communities¹¹².” This ELSI substudy seeks to implement a model called Diversity Equity and Inclusion 360 (DEI 360), which includes engaging geographically, racially-ethnically, and socio-culturally diverse stakeholders. These stakeholders include family decision-makers, tissue requesters, community advisory board members, and multi-disciplinary specialty committee members throughout the entire duration of the SMaHT Network. Feedback from community stakeholders will be leveraged to inform communication and enrollment efforts as well as dissemination of study findings.

The SMaHT Network is uniquely positioned to collaborate with many other large consortia and programs. These include: the Human Pangenome Reference Consortium (HPRC)⁹, to leverage methods for constructing haplotype-phased genome assemblies; the Impact of Genetics Variation on Function (IGVF) Consortium¹¹³, to understand the functional consequences of genetic variation; the developmental GTEx project ¹¹¹, to access datasets from tissues at early developmental stages; the Human Tumor Analysis Network (HTAN) and PreCancer Atlas, to further understand the progression from normal cells to tumor cells through somatic mutations; and PsychENCODE¹¹⁴, to inform on the phenotypic consequences of brain somatic mosaicism. These collaborations will enrich the individual studies and ultimately, through data integration and cross-network analyses, further enhance our understanding of the context and consequences of somatic mutations.

Producing the somatic mutation catalog

To produce the first phase of the somatic mutation catalog, the SMaHT Network will strike a balance between standard genomic assays, productionized and applied uniformly by the GCCs to all tissues, and bespoke assays developed by the TTD projects, focusing on novel technological approaches. As part of the initial phase of the SMaHT project, benchmarking efforts are nearing completion, using both primary human tissues and cell lines. We have used this benchmarking to determine optimal sequencing coverage, compare the accuracy of variant calling algorithms, and evaluate the utility of long- and short-read sequencing data generated on diverse sequencing platforms from multiple GCCs.

The GCCs will deploy three core assays across all tissue specimens that meet quality thresholds: deep short-read whole-genome sequencing (WGS; over 300x coverage), long-read WGS (over 30x) sequencing, and RNA sequencing (over 50 million reads). The deep short-read WGS will enable the discovery of high allele frequency somatic mutations across tissues acquired early in embryogenesis, as well as discovery of the large clonal expansions arising later in life. Since these core assays will be performed on bulk tissues, composed of heterogeneous cell types, only mutations with a relatively high variant allele frequency (VAF; above 1-2%) will be accurately detectable at the proposed depth of sequencing. The long-read WGS will facilitate the detection of complex SVs, MEIs, and variants in complex genetic loci that have been challenging to accurately study using short-read data, such as the MHC region, centromeres, telomeres, acrocentric DNA including ribosomal DNA, and other tandem-repeat regions of the genome. Ultralong-read sequencing will enable us to generate near telomere-to-telomere donor-specific reference genome assemblies (DSAs) for at least 50 donors and through reducing misalignment, enhance the discovery of diverse types of variants within an individual⁵⁹, including complex somatic SVs and other mutations in previously unmappable regions of the genome¹¹⁵. Lastly, the RNA sequencing may allow us to assess transcriptional consequences of early mutations and late clonal expansions, as well as, by comparison with single cell RNA sequencing atlases¹¹⁶, cell type composition of heterogeneous tissues.

In addition to these core assays, GCCs will deploy three approaches specifically designed to profile low-frequency somatic mutation: duplex sequencing, single cell WGS, and transcript-based detection of mutations. These technologies, while published and well-tested, represent recent innovations and have not yet been systematically deployed across sequencing centers, or applied at large scale.

As conventional DNA sequencing platforms have a non-trivial sequencing error rate (in the order of 1 in 1,000-10,000), a putative mutation needs to be detected in multiple independent reads to assure it is not artifactual. However, by sequencing both the forward and reverse strands of each individual DNA duplex molecule, this error rate is drastically reduced. Since the reduced error rate is much lower than the expected number of somatic mutations in most tissues, an average mutation burden and mutational profile can be obtained by shallow genome-wide duplex coverage (0.5-2x)^45,117. Duplex sequencing of bulk tissue samples is well-suited to finding average mutation burdens and spectra of SNVs and indels within cell populations, but the low depth generally precludes discovery of somatic CNVs and SVs, or the precise inference of variant allele frequency of specific mutations.

Even with a reduced sequencing error rate, bulk DNA sequencing will average out the mutational patterns of all cells and does not allow to assess the variability of mutational patterns between cells or the reconstruction of cell lineages. Instead, sequencing the DNA of single cells or single cell-derived clones will enable the most detailed discovery of somatic mutations. This can be achieved either by expanding single cells in vitro^3,6,25,26,43 or laser capture microdissection to isolate naturally occurring clonal populations of cells^{4,24,48,49,91}.

Alternatively, direct single cell DNA sequencing is applicable to all cell types, including non-dividing cells. However, whole-genome amplification can cause allelic or locus dropout, uneven coverage across the genome and artifactual variants introduced during biochemical amplification. The Direct Library Preparation (DLP+)^118,119 method avoids whole-genome amplification and allows for the accurate detection of CNVs at the single cell level and other mutations at the population level. The Primary Template-directed Amplification (PTA)^30,120 method offers a substantial improvement in data quality over previous single cell amplification methods, resulting in more uniform genome coverage and fewer artifactual variants. A more recent version of PTA, the ResolveOme approach, profiles both the transcriptome and the genome from the same single cell. If validated, this approach will represent a major advance in allowing new mutation detection and cellular phenotyping at the same time. Profiling somatic mutations in single cells will enable us to characterize mutational patterns and associations between mutation types and to reconstruct phylogenetic trees of normal cells across tissues. In cases of polyploid cells, the VAFs of somatic mutations may deviate from the expected 0.5 and ploidy will need to be taken in consideration in downstream analyses.

Lastly, at least some somatic mutations can be inferred from RNA^121–123. Methods that allow for the interrogation of the full-length transcriptome in single cells, such as Smart-seq3¹²⁴ or STORM-seq^125,126, can facilitate the detection of somatic mutations, such as SNVs, indels and fusion genes within transcribed regions of the genome. This allows assessing cell type specificity for clonal expansion of certain genetic variants. Furthermore, STORM-seq enables quantification of transposable element expression at single cell resolution, which has been shown to be challenging with other single-cell RNA-seq methods¹²⁵. The single cell data also provide references for a more precise deconvolution of cell types in bulk tissues.

Each of these methods for the detection of somatic mosaic variants presents its own advantages and disadvantages and thus they are complementary (Table 1). For example, while genome-wide duplex sequencing has a lower sequencing error rate and excels at population-level inferences of patterns of short mutations acquired during the entire lifespan, the low depth precludes detection of the precise allele frequency of a specific variant. Bulk sequencing at medium-high coverage (300x) will only detect variants at a sufficiently high frequency (i.e. 1-2%) in tissues, which are mostly acquired in early embryogenesis. Single cell sequencing can in principle detect all variants present in a single cell and allow reconstruction of cell phylogenies, but it requires significant costs and efforts to address genome amplification artifacts. RNA-based mutation discovery allows for direct integration of mutations with transcriptomic information but is naturally confined to expressed regions of the genome. Taken together, these genomic assays function as complementary techniques to detect somatic mutations and will enable the robust interrogation of mutational patterns across human tissues.

Table 1. Comparison between somatic mutation discovery methods.

A combination of different methods can achieve comprehensive mutation discovery and accurate analysis.

Methods	Analysis by
	Bulk sequencing	Duplex sequencing	Clonal expansion	Single cell WGA (PTA)	LCM clones
Applicability to any tissue	Yes	Yes	No	Yes	No
Mutation types discovered	All	SNVs, indels	All	All	All
Applicability to long read	Yes	Yes	Yes	Inefficient	Inefficient
Fraction of genome sampled	100%	30%-100%, depending on fragmentation method	100%	~90% per cell; 100% across cells	100%
Detection of early and clonally expanded mutations	Most	Minority	Depending on clone number	Depending on cell number	Depending on clone number
Overall mutation spectrum	No	Yes	Yes	Yes	Yes
Likely amount of artifacts	Small (10−4)	Very small (<10−8)	Small (10−4)	Some	Small (10−4)
Information on cell lineages	No	No	Yes	Yes	Yes
Advantage(s)	Sensitive detection of high frequency mutations of all types	Obtaining overall mutation spectrum even at low (0.5-2x) coverage	Accurate mutations discovery at a single cell level	Mutation discovery at a single cell level in any tissue	Accurate mutations discovery at a single cell level
Limitation(s)	Need for high coverage; missing low frequency mutations	Only SNVs & indels; missing high frequency mutations at low (0.5-2x) coverage	Applicable to culturable or reprogrammable cells	<50% sensitivity because of drop-out and amplification artifacts	Applicable to tissues with visible clonal substructures

Areas of technology development

As new technologies to interrogate somatic mutations with high resolution or sensitivity are constantly emerging, a large part of the SMaHT Network is devoted to developing new tools and technologies (Table 2). The first area of innovation aims to increase the accuracy of mutation detection in single cells or molecules by further reducing background noise. For single cell WGS, a limited cloning step to create small pools of cells can reduce allelic dropout and amplification artifacts. In parallel, the SMaHT Network aims to reduce the error rate of amplification and sequencing for single cells and molecules through various adaptations of duplex sequencing technologies^45,127–129. These approaches will allow for the interrogation of the landscape of somatic variation in single cells and complex multicellular tissues with high precision, which is crucial to study tissues without large-scale expansions.

Table 2. Experimental and analytical methods adopted by the SMaHT Network.

The SMaHT Network will apply a set of standard approaches as well as benchmark new methods to improve detection, analysis, and functional annotation of somatic variants across tissues and individuals.

Core assays (bulk)	Short-read DNA sequencing (Illumina),Short-read RNA sequencing (Illumina), Long-read DNA sequencing (PacBio, ONT) Long-read full-length transcript sequencing (PacBio)
Extended assays	Single cell DNA sequencing with Primary Template-Directed Amplification (PTA) Duplex sequencing: Concatenating Original Duplex for Error Correction (CODEC)/ Nanorate Sequencing (NanoSeq)/ Tn5-duplex-seq (CompDuplex-seq and VISTA-seq) / Ultima ppmSeq
Scale-up approaches and technology developments	Single molecule sequencing	Hairpin Duplex Enhanced Fidelity sequencing (HiDEF-seq)
	Single cell duplex sequencing	scNanoSeq and scUduplex-seq
	Structural variants detection	Single cell full length RNA-seq for MEI detection Single-cell, mini-bulk ME-targeted sequencing (PTA-HAT-seq) Single-cell Total RNA-seq Miniaturized sequencing (STORM-seq), Strand-seq
	Spatial variant detection	Slide-tags
	Genotype-to-phenotype multi-omics	Epigenome	Single cell 3C-seq protocols (e.g., Dip-C) Assay for Transposase-Accessible Chromatin (ATAC-seq), Genotyping of Targeted loci with Chromatin Accessibility (GoT-ChA, GoT-EpiM) Fiber-seq (single-molecule chromatin accessibility, nucleosome occupancy, TF occupancy, RNA polymerase occupancy, CpG methylation, and genetic variant identification) var-CUT&Tag (enrichment and epigenetic annotation of variants in regulatory elements) Strand-seq (nucleosome occupancy and SVs in the same cell)
		Transcriptome	ResolveOme method GoT-ChA-RNA Duplex-seq-based large-scale single-cell dual omics profiling
		Proteome	GoT-ChA-Pro
	Computational tools	Donor-specific reference genome assemblies (DSAs) (paired Fiber-seq, ultra-long ONT, and Hi-C data) Hybrid approach for scWGA RUFUS (a reference-free, kmer-based variant detection algorithm)
Infrastructure	Centralized three germ layer tissue collection banking with donor metadata Automated analysis pipelines for quality control A variant catalog containing a curated set of annotated somatic mutations Computational pipelines with multi-omics platforms Human pangenome visualization with somatic mutation detection pipelines A cloud-based infrastructure and a web portal

Secondly, the SMaHT Network aims to increase the sensitivity of SV detection to single molecules or cells. As SVs extend beyond the length of a typical short-read, long-read sequencing unlocks SV detection across the genome, especially for MEIs and other rearrangements in repetitive regions^130,131. However, many single cell DNA amplification approaches result in short fragments. Therefore, we are applying long-read sequencing to clonal populations such as iPSC lines which have been used²⁵ in lieu of single cell for lineage reconstructions as they can be expanded and analyzed by bulk sequencing, avoiding in vitro DNA amplification. In addition, MEIs can be cost-effectively assessed by target enrichment assays as new insertions share conserved sequences in each transposon subfamily. We are developing targeted detection of MEI insertions by utilizing Cas9-targeted long-read sequencing¹³² and PTA-amplified micro-bulk or single cells⁷⁷. These efforts will unlock the study of SVs and MEIs in all tissues and across the lifespan, even in the absence of clonal expansions.

Thirdly, the SMaHT Network will develop scalable platforms that can perform variant detection spatially in human tissues, through single cell DNA and RNA sequencing with resolved spatial barcodes^133,134. This will allow us to study the prevalence and extent of clonal expansions across ages and tissues, especially in organs without a clearly organized tissue architecture.

An outstanding question is the effect of specific somatic mutations on the phenotype of the cells that harbor them. While certain mutations are under positive selection and lead to clonal expansions, how these mutations alter cellular phenotypes is mostly unknown. The consequence of a mutation can be assessed by combining mutational readouts, either through genotyping of specific mutations^135,136 or genome sequencing, in combination with functional readouts of cells, such as the transcriptome, proteome, epigenome, methylome, and the chromatin accessibility landscape^137–142. Interpreting the phenotypic effects of somatic mutations will greatly benefit our understanding of the clinical consequences.

The efforts in tool and technology development within the SMaHT Network are focused on improving precision in somatic mutation detection and interpretation at scale, each addressing vital shortcomings of current assays, with a goal to productionize and deploy many of these within the Network at large. After the development phase, the precise extent and scope of the deployment of these assays across the SMaHT tissues and donors will depend on the cost, scalability and priorities of the Network.

Integration and analysis of data

The low VAF of mosaic variants brings unique challenges in bioinformatic analysis¹⁴³, and we expect that novel computational methods and tools are needed to fully analyze the data and to increase the sensitivity and specificity of variant detection. Somatic mutation detection algorithms developed in cancer genomics are often inadequate for detecting variants with allele fractions less than 2-5% and simply increasing the depth of sequencing is not cost-effective. Thus, more sophisticated machine learning algorithms that efficiently incorporate various local features near candidate variants may prove useful^{138–140,144}.

Other challenges include optimal integration of long-read and short-read data, inference of lineage relationships based on bulk and single cell data, and effective strategies for integrative and comparative analysis of samples across the tissues and across individuals. An important aspect of our analysis will be the use of donor-specific diploid genomes assembled using short Illumina, long PacBio and ultra-long Nanopore and Hi-C reads. Alignment to the donor-specific reference genome¹³⁷ will allow for more accurate variant identification, especially in repetitive regions, as well as for examination of allele-specific transcriptional and epigenetic modulations associated with genetic variants.

The SMaHT Data Analysis Center (DAC) will lead an effort to collect, curate, and analyze the vast amount of multi-modal data generated on multiple platforms and to create a data resource for the scientific community. The DAC will ensure high data standards with various quality control steps and compile extensive metadata describing experimental and data processing protocols, following the FAIR (Findable, Accessible, Interoperable and Reusable) guidelines¹⁴⁵. Scalable and cost-effective analytical workflows will be implemented on a cloud platform with full provenance and docker images to enable reproducibility of the analysis output.

The data generated by the consortium will be made available to the wider scientific community via a user-friendly and secure web portal (https://data.smaht.org). This portal will feature: (i) a reference catalog of somatic variants that can be searched (e.g., by locus, tissue, or phenotypic features such as age) and annotated with information from other genomics databases; (ii) a workbench that enables users to apply the computational pipelines developed by the SMaHT Network to their own data, and (iii) data visualization tools including a multi-scale browser that allows users to navigate the data from a genome-level view to the sequencing read-level view. Visual inspection of variants using such a browser will be particularly helpful in assessing their quality, and the annotations will enable rapid identification of variants that may be functionally relevant.

Conclusion

The SMaHT Network aims to produce a comprehensive reference catalog of somatic mutations, across tissues and individuals, by harnessing the full potential of many different genomic assays, including short and long-read bulk WGS, duplex sequencing, ultralong read sequencing, single cell DNA sequencing, and RNA sequencing (Fig. 3). The Network will develop new tools and technologies to increase our ability to detect somatic mutations as well as infer their phenotypic consequences at greater resolution. All these various data modalities will be integrated, analyzed, and released to the research community and wider public.

Fig. 3 |. Methods, assays and questions:

Overview of sampling methods and sequencing assays deployed in the SMaHT Network, as well as the biological questions, outcomes and inferred mutational patterns from downstream analyses of the catalog of somatic mutations across normal tissues, including mutation rates or burdens, selection, lineage tracing and mutational signatures (reference signatures obtained from https://cancer.sanger.ac.uk/signatures)⁵⁵.

An extensive catalog of somatic mutations will reveal mutational patterns, rates, and signatures across tissues, allowing us to infer the biological and molecular processes that govern somatic mutagenesis and their adaptive and maladaptive consequences for development and disease (Fig. 3). Our assays can inform on mutations under selection in tissues, which result in clonal expansions and potentially tissue dysfunction. Single cell analyses added to the bulk readouts will further allow us to generate cellular phylogenies of human development, infer embryonic differentiation dynamics and improve our future assessment of de novo germline mutations.

Delineating the full extent of somatic mosaicism greatly exceeds the scope of the Human Genome Project. A typical cell may acquire hundreds to thousands of somatic mutations in a lifetime. There are trillions of cells in a human body and so the total number of somatic mutations acquired in a single individual may well exceed quadrillions, millions of times the size of the human genome. Beyond cataloging somatic variation across tissues, it is essential to understand the causes, patterns, and consequences of somatic mutations in normal cells, and provide a crucial comparison baseline for disease research. The efforts of the SMaHT Network will substantially contribute to our insights into the role of somatic variation in health, aging, and disease.

Participants in the Somatic Mosaicism across Human Tissues Network

National Institutes of Health

Richard S. Conroy⁶⁹, Brionna Y. Hair⁶⁹, Walter J. Koroshetz⁷⁰, Roger Little⁷¹, Amy C. Lossie^71*, Jill A. Morris⁷⁰, Dena C. Procaccini⁶⁹, Wendy Wang⁷²

*Dr. Lossie was substantially involved in UM1DA058219, UM1DA058220, UM1DA058229, UM1DA058230, UM1DA058235, and UM1DA058236, consistent with his/her role as Program Officer. She is also the NIH Working Group Coordinator, and has involvement with the remaining awards, consistent with this role.

Organizational Center (U24NS132103)

Ting Wang^8,43*, Casey Andrews⁸, Lucinda Antonacci-Fulton⁸, Sarah Cody⁸, Milinn Kremitzki⁸, Heather A Lawson⁴³, Daofeng Li⁴³, Tina Lindsay⁸, Wenjin Zhang⁴³

*Contact: twang@wustl.edu

Tissue Procurement Center (U24MH133204)

Thomas J. Bell^12*, Thomas G. Blanchard¹⁷, Valerie J. Estela-Pro¹², Melissa A. Faith^28,29, Kayla Giancarlo⁷³, Melissa Grimm⁷³, Azra Hasan¹², Raquel G. Hernandez3^5,36, M. Kathryn Leonard¹², Phoebe McDermott²⁹, Mary Pfeiffer⁷³, Isabel Sleeman¹², Melissa W. VonDran¹²

*Contact: tbell@ndriresource.org

Genome Characterization Centers

University of Washington – James Bennett (UM1DA058220)

James T Bennett^13,14*, Stephanie Bohaczuk²⁴, Colleen P Davis²⁴, Evan E Eichler^24,25, Chris Frazar²⁴, Kendra Hoekzema²⁴, Meng-Fan Huang²⁴, Caitlin Jacques²⁴, Dana M. Jensen¹³, Tom Kolar²⁴, Youngjun Kwon²⁴, Kelsey Loy¹³, Yizi Mao²⁴, Sohn Min-Hwan²⁴, Katherine M Munson²⁴, Shane Neph²⁴, Jeffrey Ou¹³, Nancy L Parmalee¹³, Minh-Hang M Pham¹³, Jane Ranchalis⁵⁹, Luyao Ren²⁴, Adriana E Sedeño-Cortés⁵, Josh Smith²⁴, Melanie Sorensen²⁴, Andrew B Stergachis^24,58,59, Lila Sutherlin¹³, Mitchell R Vollger²⁴, Chia-Lin Wei²⁴, Jeffrey M Weiss²⁴, Christina Zakarian²⁴

*Contact: jtbenn@uw.edu

Genome Characterization Center

Baylor College of Medicine – Richard Gibbs (UM1DA058229)

Richard A. Gibbs^22,23*, Elizabeth G. Atkinson²³, Sravya Bhamidipati²², Hsu Chao²², Rui Chen^22,23, Christopher M. Grochowski²², Harsha Doddapaneni²², Divya Kalra²², Ziad Khan2², Kavya Kottapalli²², Marie-Claude Gingras²², Walker Hale²², Heer Mehta²², Donna M. Muzny²², Muchun Niu²³, Luis Paulin²³, Jeffrey Rogers²², Evette Scott²², Fritz J. Sedlazeck^22,23, Kimberly Walker²², Tao Wu²³, Chenghang Zong²³

*Contact: agibbs@bcm.edu

Genome Characterization Center

Broad Institute of MIT and Harvard – Kristin Ardlie (UM1DA058235)

Kristin G. Ardlie¹, Viktor Adalsteinsson¹, Lisa Anderson¹, Carrie Cibulskis¹, Tim H. H. Coorens¹, Laura Domènech¹, Kiran Garimella¹, Whitney Hornsby¹, Steve Huang¹, Satoshi Koyama¹, Niall Lennon¹, Stephen B. Montgomery^48,49,50, Tetsushi Nakao¹, Azeet Narayan¹, Pradeep Natarajan^1,51,52, Evin Padhi⁴⁹, Constantijn Scharlee¹, Md Mesbah Uddin^1,51, Liying Xue¹, Zhi Yu¹, Shadi Zaheri¹

*Contact: kardlie@broadinstitute.org

Genome Characterization Center

Washington University – Ting Wang (UM1DA058219)

Ting Wang^8,43*, Derek Albracht⁸, Eddie Belter⁸, Emma Casey⁸, Justin Chen⁴³, Yuchen Cheng⁸, Shihua Dong⁸, Qichen Fu, Robert Fulton⁸, John Garza, H. Josh Jang⁵⁷, Sheng Chih Jin⁸, Benjamin K. Johnson⁵⁷, Nahyun Kong⁸, Daofeng Li⁵⁷, Vivien Li⁸, Tina Lindsay⁸, Shane Liu⁸, Juan Macias-Velasco⁸, Elvisa Mehinovic⁸, Benpeng Miao⁸, Theron Palmer⁵⁷, Purva Patel⁸, Mary Rhodes⁵⁷, Dan Rohrer⁵⁷, Andrew Ruttenberg⁸, Ayush Semwal⁵⁷, Hui Shen⁵⁷, Jiawei Shen⁸, Zitian Tang⁸, Chad Tomlinson⁸, Wenjin Zhang⁸, Xin Zilan⁸

*Contact: twang@wustl.edu

Genome Characterization Center

New York Genome Center – Soren Germer (UM1DA058236)

Soren Germer^11*, Samuel Aparicio^9,10,11, Jade E. B. Carter¹¹, Bill Driscoll¹¹, Uday Evani¹¹, Heather Geiger¹¹, Tausif Hasan¹¹, Manisha Kher¹¹, Dan A. Landau^11,41,42, Rajeeva Musunuri¹¹, Giuseppe Narzisi¹¹, Nicolas Robine¹¹, Alexi Runnels¹¹

*Contact: sgermer@nygenome.org

Data Analysis Center (UM1DA058230)

Peter J. Park^33,53*, Mingyun Bae^4,5, Michele Berselli³³, Ann Caplin³³, Hye-Jung Elizabeth Chun³³, Niklas Engel³³, William Feng³³, Yan Gao³³, Nils Gehlenborg³³, Dominik Glodzik³³, Yoo-Jin Ha³³, Hu Jin³³, Sehi ĽYi³³, Eunjung Alice Lee^1,4,5,6, Heng Li^33,44, Po-Ru Loh^1,45, Lovelace J. Luquette³³, Maximilian Marin^33,44, Julia Markowski³³, Dominika Maziec³³, Huang Neng^33,44, Sarah Nicholson³³, Junseok Park^4,5, Qin Qian⁴⁴, Han Flora Qu³³, Douglas Rioux³³, William Ronchetti³³, Andrew Schroeder³³, Corinne Sexton3³, Yichen Si, Kar-Tong Tan⁴⁴, David Tang⁴⁵, Alexander D. Veit³³, Vinayak V. Viswanadham³³, Suenghyun Wang³³, Kate Woo³³, Xi Zeng^4,5, Yuwei Zhang³³, Yifan Zhao³³, Ying Zhou⁴⁴

*Contact: peter_park@hms.harvard.edu

Tool and Technology Development

University of Michigan – Ryan Mills (UG3NS132084)

Ryan E. Mills^18,19*, Brandt A. Bessell¹⁸, Alan P. Boyle^18,19, Ingrid Flashpohler¹⁸, Steve Losh¹⁸, Michael J McConnell⁴⁷, Torrin L. McDonald¹⁸, Camille Mumm¹⁹, Jessica A. Switzenberg¹⁸, Jinhao Wang¹⁸, Weichen Zhou¹⁸

*Contact: remills@med.umich.edu

Tool and Technology Development

Boston’s Children Hospital – Sangita Choudhury (UG3NS132144)

Sangita Choudhury^1,4,6*, Guanlan Dong^1,4,6, Nazia Hilal^1,4,6, Se-Young Jo^4,6, Eunjung Alice Lee^1,4,5,6, Shayna L Mallett^1,4,6, Monica Devi Manam^4,6, Shulin Mao^4,6, Diane D. Shao^4,6,56, Christopher Walsh^4,6,63, Sijing Zhao^4,6

*Contact: Sangita.Choudhury@childrens.harvard.edu

Tool and Technology Development

Stanford University – Alexander Urban (UG3NS132146)

Alexander E Urban^49,62*, Yiling Elaine Huang⁴⁹, Jan O. Korbel^39,40, Abhiram Natu⁶⁶, Reenal Pattni⁴⁹, Carolin Purman⁴⁹, Flora M. Vaccarino^66,67,68, Bo Zhou⁴⁹, Xiaowei Zhu⁷⁴

*Contact: aeurban@stanford.edu

Tool and Technology Development

Baylor College of Medicine – Chenghang Zong (UG3NS132132)

Chenghang Zong^23*, Jiayi Luo, Muchun Niu²³, Yichi Niu²³, Rohan Thakur⁷⁵, David A. Weitz⁷⁵, Yang Zhang²³

*Contact: Chenghang.Zong@bcm.edu

Tool and Technology Development

Baylor College of Medicine – Fritz Sedlazeck (UG3NS132105)

Fritz J. Sedlazeck^22,23,55*, Yilei Fu²², Michal B. Izydorczyk²², Luis F. Paulin²², Tao P. Wu^23,64,65, Xiaomei Zhan²², Xinchang Zheng²²

*Contact: Fritz.Sedlazeck@bcm.edu

Tool and Technology Development

Mayo Clinic – Alexej Abyzov (UG3NS132128)

Alexej Abyzov^7*, Geon Hue Bae^2,3, Taejeong Bae⁷, Areum Cho^2,3, June Hyug Choi^2,3, Yujin Angelina Choi², Hyungbin Chun^2,3, Mrunal Dehankar⁷, Yeonjun Jang⁷, Seok-Won Jeong^2,3, Min Ji³, Mee Sook Jun^3,4, Su Rim Kim^2,3, Seong Gyu Kwon^2,3, Soung-Hoon Lee^2,3, Nam Seop Lim^2,3, Nanda Maya Mali^2,3, Ji Won Oh^2,3, Arijit Panda⁷, Jung Min Park^2,3, JaeEun Shin^2,3, Milovan Suvakov⁷

*Contact: Abyzov.Alexej@mayo.edu

Tool and Technology Development

University of Utah – Gabor Marth (UG3NS132134)

Gabor T. Marth^46*, Brad Demarest⁴⁶, Stephanie Gardiner⁴⁶, Stephanie J. Georges⁴⁶, Hunter R. Underhill^60,61, Yingqi Zhang⁶⁰

*Contact: gmarth@genetics.utah.edu

Tool and Technology Development

Weill Cornell Medicine – Dan Landau (UG3NS132139)

Dan A. Landau^11,41,42*, Samantha Avaylon¹¹, Alexandre Cheng^11,41, Wei-Yu Chi⁴¹, Mariela Cortés-Lopéz⁴¹, Andrew R. D’Avino^11,41, Husain Danish^11,41, Elliot Eton^11,41, Foteini Fotopoulo^11,41, Saravanan Ganesan^11,2, Yiyun Lin^11,41, Qing Luo^11,41, Levan Mekerishvili^11,41, Joe Pelt^11,41, Catherine Potenski^11,41, Tamara Prieto^11,41, Jake Qiu^11,41, Ivan Raimondi^11,2, Rahul Satija^11,54, Manu Singh^11,41, Dennis Yuan^11,41, John Zinno^11,41

*Contact: dlandau@nygenome.org

Tool and Technology Development

New York University – Gilad Evrony (UG3NS132024)

Gilad D. Evrony^26,27*, Benjamin Costa^26,27, Jonathan Evan Shoag^79,80, Jimin Tan⁷⁶, Aristotelis Tsirigos^77,78

*Contact: Gilad.Evrony@nyulangone.org

Tool and Technology Development

Boston’s Children Hospital – Christopher Walsh (UG3NS132138)

Christopher A. Walsh^4,6,63*, Emre Caglayan^4,6,, Hayley Cline^4,6,, Niklas L. Engel³³, Shelbi E. Gill^4,6,, Robert Sean Hill^4,6,, Andrea J. Kriz^4,6,, Julia Markowski¹, Alisa Mo^4,6,, Peter J. Park^33,53, Daniel Snellings^4,6,, Vinayak V. Viswanadham³³

*Contact: Christopher.Walsh@childrens.harvard.edu

Tool and Technology Development

Dana Farber Cancer Institute – Kathleen Burns (UG3NS132127)

Kathleen H. Burns^1,16,21*, Justin S. Becker^1,15,16, Bradley E. Bernstein^1,15,16, Aidan H. Burn^1,16,21, Wen-Chih Cheng^1,16,21, Jennifer A. Karlow^1,16,21, Cheuk-Ting Law^1,16,21, Eunjung Alice Lee^1,4,5,6, Shayna L. Mallet^1,4,5, Carlos Mendez-Dorantes^1,16,21, Khue H. Nguyen^1,4,5, Adam Voshall^1,4,5, Boxun Zhao^1,4,5

*Contact: KathleenH_Burns@DFCI.HARVARD.EDU

Tool and Technology Development

Broad Institute of MIT and Harvard – Fei Chen (UG3NS132135)

Fei Chen^1,20*, Jason D. Buenrostro^1,20, Tim H.H. Coorens¹, Gad Getz^1,16,34, Benno Orr¹, Andrew J.C. Russell¹

*Contact: chenf@broadinstitute.org

Tool and Technology Development

Case Western University – Fulai Jin (UG3NS132061)

Fulai Jin^37,38*, He Li^37,38, Yan Li^37,38, Leina Lu^37,38, Xiaofeng Zhu⁸¹

*Contact: fxj45@case.edu

Tool and Technology Development

University of Massachusetts – Thomas Fazzio (UG3NS132136)

Thomas G. Fazzio^30*, Trishita Basak³⁰, Manuel Garber³², Azita Ghodssi³⁰, Katrina Newcomer³⁰, Yuqing Wang³²

*Contact: Thomas.Fazzio@umassmed.edu

Affiliations (continued from main list)

69. Office of Strategic Coordination, Division of Program Coordination, Planning, and Strategic Initiatives, National Institutes of Health, Bethesda, MD, USA

70. National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA

71. Division of Neuroscience and Behavior, National Institute on Drug Abuse, National Institutes of Health, Bethesda, MD, USA

72. Division of Cancer Prevention, National Cancer Institute, Bethesda, MD, USA

73. ConnectLife, Williamsville, NY, USA

74. City University of Hong Kong, Department of Neuroscience, City University of Hong Kong, Hong Kong, China

75. Department of Physics, Harvard University, Boston, MA, USA

76. Institute for Systems Genetics, NYU Grossman School of Medicine, New York, NY, USA

77. Department of Medicine, NYU Grossman School of Medicine, New York, NY, USA

78. Department of Pathology, NYU Grossman School of Medicine, New York, NY, USA

79. Department of Urology, University Hospitals Cleveland Medical Center, Case Western Reserve University School of Medicine, Cleveland, OH, USA

80. Case Comprehensive Cancer Center, Cleveland, OH, USA

81. Department of Population and Quantitative Health Sciences, Case Western Reserve University, Cleveland, OH, USA