Search

The coronavirus disease 2019 (COVID-19) pandemic has caused millions of deaths around the world and revealed the need for data-driven models of pandemic spread. Accurate pandemic caseload forecasting allows informed policy decisions on the adoption of non-pharmaceutical interventions (NPIs) to reduce disease transmission. Using COVID-19 as an example, we present Pandemic conditional Ordinary Differential Equation (PAN-cODE), a deep learning method to forecast daily increases in pandemic infections and deaths. By using a deep conditional latent variable model, PAN-cODE can generate alternative caseload trajectories based on alternate adoptions of NPIs, allowing stakeholders to make policy decisions in an informed manner. PAN-cODE also allows caseload estimation for regions that are unseen during model training. We demonstrate that, despite using less detailed data and having fully automated training, PAN-cODE’s performance is comparable to state-of-the-art methods on 4-week-ahead and 6-week-ahead forecasting. Finally, we highlight the ability of PAN-cODE to generate realistic alternative outcome trajectories on select US regions.

Recent literature has highlighted the role of the host in the prognosis of oral squamous cell carcinoma (OSCC). In this study, we retrospectively examined the impact of autoimmune (AI) disorders as an aspect of the host status on survival outcomes in OSCC patients.From a departmental database of OSCC patients (n = 1369), 123 patients with an AI disorder were identified. AI and no-AI groups were compared for survival outcomes.There were no significant differences in survival between groups for overall survival, disease-specific survival, local, regional, and distant recurrence-free probabilities. However, survival and recurrence-free probabilities were poorer in the AI group versus the no AI group.Patients with AI disease trended towards worse outcomes. This suggests immune dysregulation in these patients may impact oncologic outcomes.

Supplementary Data from Reconstructing Complex Cancer Evolutionary Histories from Multiple Bulk DNA Samples Using Pairtree

Cancers are composed of genetically distinct subpopulations of malignant cells. DNA-sequencing data can be used to determine the somatic point mutations specific to each population and build clone trees describing the evolutionary relationships between them. These clone trees can reveal critical points in disease development and inform treatment. Pairtree is a new method that constructs more accurate and detailed clone trees than previously possible using variant allele frequency data from one or more bulk cancer samples. It does so by first building a Pairs Tensor that captures the evolutionary relationships between pairs of subpopulations, and then it uses these relations to constrain clone trees and infer violations of the infinite sites assumption. Pairtree can accurately build clone trees using up to 100 samples per cancer that contain 30 or more subclonal populations. On 14 B-progenitor acute lymphoblastic leukemias, Pairtree replicates or improves upon expert-derived clone tree reconstructions.Significance:Clone trees illustrate the evolutionary history of a cancer and can provide insights into how the disease changed through time (e.g., between diagnosis and relapse). Pairtree uses DNA-sequencing data from many samples of the same cancer to build more detailed and accurate clone trees than previously possible.See related commentary by Miller, p. 176.This article is highlighted in the In This Issue feature, p. 171.

Targeted-sequencing analysis of PDX and PairTree predicted mutational population clusters

dPDX and dRI-PDX leukemia-initiating cell frequencies

Disease recurrence causes significant mortality in B-progenitor acute lymphoblastic leukemia (B-ALL). Genomic analysis of matched diagnosis and relapse samples shows relapse often arising from minor diagnosis subclones. However, why therapy eradicates some subclones while others survive and progress to relapse remains obscure. Elucidation of mechanisms underlying these differing fates requires functional analysis of isolated subclones. Here, large-scale limiting dilution xenografting of diagnosis and relapse samples, combined with targeted sequencing, identified and isolated minor diagnosis subclones that initiate an evolutionary trajectory toward relapse [termed diagnosis Relapse Initiating clones (dRI)]. Compared with other diagnosis subclones, dRIs were drug-tolerant with distinct engraftment and metabolic properties. Transcriptionally, dRIs displayed enrichment for chromatin remodeling, mitochondrial metabolism, proteostasis programs, and an increase in stemness pathways. The isolation and characterization of dRI subclones reveals new avenues for eradicating dRI cells by targeting their distinct metabolic and transcriptional pathways before further evolution renders them fully therapy-resistant.Significance:Isolation and characterization of subclones from diagnosis samples of patients with B-ALL who relapsed showed that relapse-fated subclones had increased drug tolerance and distinct metabolic and survival transcriptional programs compared with other diagnosis subclones. This study provides strategies to identify and target clinically relevant subclones before further evolution toward relapse.See related video: https://vimeo.com/442838617See related article by E. Waanders et al.

Patient characteristics and patient genomic analysis (WES and SNP 6.0)

Leukemia-initiating cell frequencies of paired diagnosis and relapse patient samples

RNA-sequencing data from Patient 9 PDX

RNA-sequencing, pathway enrichment (GSEA) reports and GSVA results (including gene list HSC vs B) of PDX and paired patient samples

PDX targeted-sequencing tissue concordance

Supplementary Data from Activating mTOR Mutations Are Detrimental in Nutrient-Poor Conditions

Thousands of RNA-binding proteins (RBPs) crosslink to cellular mRNA. Among these are numerous unconventional RBPs (ucRBPs)—proteins that associate with RNA but lack known RNA-binding domains (RBDs). The vast majority of ucRBPs have uncharacterized RNA-binding specificities. We analyzed 492 human ucRBPs for intrinsic RNA-binding in vitro and identified 23 that bind specific RNA sequences. Most (17/23), including 8 ribosomal proteins, were previously associated with RNA-related function. We identified the RBDs responsible for sequence-specific RNA-binding for several of these 23 ucRBPs and surveyed whether corresponding domains from homologous proteins also display RNA sequence specificity. CCHC-zf domains from seven human proteins recognized specific RNA motifs, indicating that this is a major class of RBD. For Nudix, HABP4, TPR, RanBP2-zf, and L7Ae domains, however, only isolated members or closely related homologs yielded motifs, consistent with RNA-binding as a derived function. The lack of sequence specificity for most ucRBPs is striking, and we suggest that many may function analogously to chromatin factors, which often crosslink efficiently to cellular DNA, presumably via indirect recruitment. Finally, we show that ucRBPs tend to be highly abundant proteins and suggest their identification in RNA interactome capture studies could also result from weak nonspecific interactions with RNA.

Cancers are composed of genetically distinct subpopulations of malignant cells. DNA-sequencing data can be used to determine the somatic point mutations specific to each population and build clone trees describing the evolutionary relationships between them. These clone trees can reveal critical points in disease development and inform treatment. Pairtree is a new method that constructs more accurate and detailed clone trees than previously possible using variant allele frequency data from one or more bulk cancer samples. It does so by first building a Pairs Tensor that captures the evolutionary relationships between pairs of subpopulations, and then it uses these relations to constrain clone trees and infer violations of the infinite sites assumption. Pairtree can accurately build clone trees using up to 100 samples per cancer that contain 30 or more subclonal populations. On 14 B-progenitor acute lymphoblastic leukemias, Pairtree replicates or improves upon expert-derived clone tree reconstructions. Significance: Clone trees illustrate the evolutionary history of a cancer and can provide insights into how the disease changed through time (e.g., between diagnosis and relapse). Pairtree uses DNA-sequencing data from many samples of the same cancer to build more detailed and accurate clone trees than previously possible. See related commentary by Miller, p. 176. This article is highlighted in the In This Issue feature, p. 171.

Our ability to understand the genotype-to-phenotype relationship is hindered by the lack of detailed understanding of phenotypes at a single-cell level. To systematically assess cell-to-cell phenotypic variability, we combined automated yeast genetics, high-content screening and neural network-based image analysis of single cells, focussing on genes that influence the architecture of four subcellular compartments of the endocytic pathway as a model system. Our unbiased assessment of the morphology of these compartments—endocytic patch, actin patch, late endosome and vacuole—identified 17 distinct mutant phenotypes associated with ~1,600 genes (~30% of all yeast genes). Approximately half of these mutants exhibited multiple phenotypes, highlighting the extent of morphological pleiotropy. Quantitative analysis also revealed that incomplete penetrance was prevalent, with the majority of mutants exhibiting substantial variability in phenotype at the single-cell level. Our single-cell analysis enabled exploration of factors that contribute to incomplete penetrance and cellular heterogeneity, including replicative age, organelle inheritance and response to stress.

About 15% of human cancer cases are attributed to viral infections. To date, virus expression in tumor tissues has been mostly studied by aligning tumor RNA sequencing reads to databases of known viruses. To allow identification of divergent viruses and rapid characterization of the tumor virome, we develop viRNAtrap, an alignment-free pipeline to identify viral reads and assemble viral contigs. We utilize viRNAtrap, which is based on a deep learning model trained to discriminate viral RNAseq reads, to explore viral expression in cancers and apply it to 14 cancer types from The Cancer Genome Atlas (TCGA). Using viRNAtrap, we uncover expression of unexpected and divergent viruses that have not previously been implicated in cancer and disclose human endogenous viruses whose expression is associated with poor overall survival. The viRNAtrap pipeline provides a way forward to study viral infections associated with different clinical conditions.

About 15% of human cancer cases are attributed to viral infections. To date, virus expression in tumor tissues has been mostly studied by aligning tumor RNA sequencing reads to databases of known viruses. To allow identification of divergent viruses and rapid characterization of the tumor virome, we developed viRNAtrap, an alignment-free pipeline to identify viral reads and assemble viral contigs. We apply viRNAtrap, which is based on a deep learning model trained to discriminate viral RNAseq reads, to 14 cancer types from The Cancer Genome Atlas (TCGA). We find that expression of exogenous cancer viruses is associated with better overall survival. In contrast, expression of human endogenous viruses is associated with worse overall survival. Using viRNAtrap, we uncover expression of unexpected and divergent viruses that have not previously been implicated in cancer. The viRNAtrap pipeline provides a way forward to study viral infections associated with different clinical conditions.

Modelling both primary sequence and secondary structure preferences for RNA binding proteins (RBPs) remains an ongoing challenge. Current models use varied RNA structure representations and can be difficult to interpret and evaluate. To address these issues, we present a universal RNA motif-finding/scanning strategy, termed PRIESSTESS (Predictive RBP-RNA InterpretablE Sequence-Structure moTif regrESSion), that can be applied to diverse RNA binding datasets. PRIESSTESS identifies dozens of enriched RNA sequence and/or structure motifs that are subsequently reduced to a set of core motifs by logistic regression with LASSO regularization. Importantly, these core motifs are easily visualized and interpreted, and provide a measure of RBP secondary structure specificity. We used PRIESSTESS to interrogate new HTR-SELEX data for 23 RBPs with diverse RNA binding modes and captured known primary sequence and secondary structure preferences for each. Moreover, when applying PRIESSTESS to 144 RBPs across 202 RNA binding datasets, 75% showed an RNA secondary structure preference but only 10% had a preference besides unpaired bases, suggesting that most RBPs simply recognize the accessibility of primary sequences.

Cancer genomes harbor a catalog of somatic mutations. The type and genomic context of these mutations depend on their causes, and allow their attribution to particular mutational signatures. Previous work has shown that mutational signature activities change over the course of tumor development, but investigations of genomic region variability in mutational signatures have been limited. Here, we expand upon this work by constructing regional profiles of mutational signature activities over 2,203 whole genomes across 25 tumor types, using data aggregated by the Pan-Cancer Analysis of Whole Genomes (PCAWG) consortium. We present GenomeTrackSig as an extension to the TrackSig R package to construct regional signature profiles using optimal segmentation and the expectation-maximization (EM) algorithm. We find that 426 genomes from 20 tumor types display at least one change in mutational signature activities (changepoint), and 257 genomes contain at least one of 54 recurrent changepoints shared by seven or more genomes of the same tumor type. Five recurrent changepoint locations are shared by multiple tumor types. Within these regions, the particular signature changes are often consistent across samples of the same type and some, but not all, are characterized by signatures associated with subclonal expansion. The changepoints we found cannot strictly be explained by gene density, mutation density, or cell-of-origin chromatin state. We hypothesize that they reflect a confluence of factors including evolutionary timing of mutational processes, regional differences in somatic mutation rate, large-scale changes in chromatin state that may be tissue type-specific, and changes in chromatin accessibility during subclonal expansion. These results provide insight into the regional effects of DNA damage and repair processes, and may help us localize genomic and epigenomic changes that occur during cancer development.

About 15% of human cancer cases are attributed to viral infections. To date, virus expression in tumor tissues has been mostly studied by aligning tumor RNA sequencing reads to databases of known viruses. To allow identification of divergent viruses and rapid characterization of the tumor virome, we develop viRNAtrap, an alignment-free pipeline to identify viral reads and assemble viral contigs. We utilize viRNAtrap, which is based on a deep learning model trained to discriminate viral RNAseq reads, to explore viral expression in cancers and apply it to 14 cancer types from The Cancer Genome Atlas (TCGA). Using viRNAtrap, we uncover expression of unexpected and divergent viruses that have not previously been implicated in cancer and disclose human endogenous viruses whose expression is associated with poor overall survival. The viRNAtrap pipeline provides a way forward to study viral infections associated with different clinical conditions.

Pairtree is a clone tree reconstruction algorithm that uses somatic point mutations to build clone trees describing the evolutionary history of individual cancers. Using the Pairtree software package, we describe steps to preprocess somatic mutation data, cluster mutations into subclones, search for clone trees, and visualize clone trees. Pairtree builds clone trees using up to 100 samples from a single cancer with at least 30 subclonal populations. For complete details on the use and execution of this protocol, please refer to Wintersinger et al. (2022).

Background: Knowledge of a patient’s tumor type is essential for guiding clinical treatment decisions in cancer, but histologically-based diagnosis remains challenging for a subset of cancers. Genomic alterations are highly indicative of tumor type and can be used to build classifiers that predict diagnoses, but most genomic-based classification methods use whole genome sequencing (WGS) data which is not feasible for widespread clinical implementation at present. Clinical sequencing is typically performed using cancer gene panels that target individual mutations, often drivers, but previous tumor type classifiers developed using driver-based features alone perform poorly. We hypothesize that a classifier developed using state-of-the-art deep-learning methods and a sufficiently large training cohort would be able to overcome previous accuracy limitations and support the development of a clinically-relevant tumor type prediction model. Methods: We present Deep Genome-Derived Diagnosis (GDD-ENS), an ensemble-based deep-learning tumor type classification method trained using data from cancer gene panel sequencing. We specifically use data from MSK-IMPACT, an FDA-authorized clinical sequencing assay that reports genomic alterations including mutations, indels, copy number alterations, and gene fusions across 505 cancer-associated genes. We aggregated a discovery cohort of 35,372 patients with solid tumors profiled with MSK-IMPACT across 38 common cancer types and used this set to generate 4,487 somatic mutation features for development. Results: GDD-ENS achieves 78.8% accuracy on a held out validation cohort of 6971 patients. For the 71.9% of predictions assigned a high confidence by the model, accuracy increases to 92.7%, rivaling WGS-based models. We use Shapley Values to report prediction-specific feature importance, and aggregate them across cancer types to show GDD-ENS identifies known cancer type-genomic alteration trends. GDD-ENS also, with high accuracy, identifies patients with cancer types not included in the 38 common types using metrics derived from ensemble statistics. For patients where non-genomic information could further guide predictions, we implement a customizable prediction-specific adaptive prior distribution and report improved accuracy after adjusting predictions to account for features such as metastatic sample biopsy site. Finally, we apply GDD-ENS to a set of 1,123 patients with Cancers of Unknown Primary (CUP) and return high confidence predictions for 49% of these patients, in some cases matching predictions on CUP samples with diagnoses that were later confirmed through additional sampling and disease progression. Conclusions: Integrating GDD-ENS into prospective clinical sequencing workflows will enable clinically-relevant tumor type predictions that can guide treatment decisions in real-time. Citation Format: Madison Darmofal, Shalabh Suman, Gurnit Atwal, Jie-Fu Chen, Anna Varghese, Jason C. Chang, Anoop Balakrishnan Rema, Aijazuddin Syed, Quaid Morris, Michael Berger. Deep-learning model for tumor type classification enables enhanced clinical decision support in cancer diagnosis. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5440.

Cancer patients, particularly those receiving B cell-depleting therapy for lymphoid malignancies, are at risk of prolonged SARS-CoV-2 infection, poorer clinical outcomes, and delayed initiation or disruption of cancer-directed therapy (Lee at al., 2022, Clark et al., 2021). We first studied T-cell mediated response to the Wuhan strain of SARS-CoV-2 in a cohort of 69 patients with hematologic and solid cancers, including 18 patients who received prior B-cell depleting therapy. Patients with prolonged COVID-19 clearance, defined by a positive PCR test for longer than 30 days, had a broad but poorly converged CD8+ dominant response and a lacking CD4+ response. To conduct this analysis, we performed bulk T-cell receptor (TCR) sequencing of 121 blood samples and tracked over time TCR repertoire statistics such as clonality, convergence, breadth, and depth of COVID-19-associated TCRs during the active and convalescent periods of COVID-19 infection. These SARS-CoV-2-associated TCRs were identified leveraging immunoSEQ T-MAP database (Snyder et al., 2020), a set of TCR sequences derived from COVID-19 patients and experimentally identified as responsive to MHC Class I and II epitopes from the Wuhan SARS-CoV-2 strain using the multiplex identification of TCR antigen assay (Klinger et al., 2015). To extend our TCR repertoire analysis to other SARS-CoV-2 variants, including Omicron, we developed a deep learning (DL) method to predict TCR specificities for new SARS-CoV-2 epitopes. This DL approach also permits the identification of SARS-CoV2-responsive TCRs private to an individual. Combining this DL approach with our TCR statistics methodology, we studied the dynamics of T-cell response to COVID-19 vaccinations in a cohort of 50 patients with cancer and analyzed TCR repertoire characteristics associated with different degrees of COVID-19 severity in a cohort of 42 cancer patients who contracted the Omicron. Understanding cellular response to novel infections is critical for patient care in the context of cancer, and our novel DL-based approach can leverage existing datasets to analyze and track response to emerging viral strains. Citation Format: Olga Lyudovyk, Yuval Elhanati, Artem Streltsov, Quaid Morris, Santosha Vardhana, Benjamin Greenbaum. T-cell mediated response to emerging COVID-19 strains in patients with cancer studied via deep learning [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 795.

The mTOR is a key regulator of cell growth that integrates growth factor signaling and nutrient availability and is a downstream effector of oncogenic receptor tyrosine kinases (RTK) and PI3K/Akt signaling. Thus, activating mTOR mutations would be expected to enhance growth in many tumor types. However, tumor sequencing data have shown that mTOR mutations are enriched only in renal clear cell carcinoma, a clinically hypervascular tumor unlikely to be constrained by nutrient availability. To further define this cancer-type–specific restriction, we studied activating mutations in mTOR. All mTOR mutants tested enhanced growth in a cell-type agnostic manner under nutrient-replete conditions but were detrimental to cell survival in nutrient-poor conditions. Consistently, analysis of tumor data demonstrated that oncogenic mutations in the nutrient-sensing arm of the mTOR pathway display a similar phenotype and were exceedingly rare in human cancers of all types. Together, these data suggest that maintaining the ability to turn off mTOR signaling in response to changing nutrient availability is retained in most naturally occurring tumors. Significance: This study suggests that cells need to inactivate mTOR to survive nutrient stress, which could explain the rarity of mTOR mutations and the limited clinical activity of mTOR inhibitors in cancer.

Cancer results from an evolutionary process that yields a heterogeneous tumor with distinct subpopulations and varying sets of somatic mutations. This perspective discusses computational methods to infer models of evolutionary processes in cancer that aim to improve our understanding of tumorigenesis and ultimately enhance current clinical practice.

Understanding the regulatory interactions that control gene expression during the development of novel tissues is a key goal of evolutionary developmental biology. Here, we show that Mbnl3 has undergone a striking process of evolutionary specialization in eutherian mammals resulting in the emergence of a novel placental function for the gene. Mbnl3 belongs to a family of RNA-binding proteins whose members regulate multiple aspects of RNA metabolism. We find that, in eutherians, while both Mbnl3 and its paralog Mbnl2 are strongly expressed in placenta, Mbnl3 expression has been lost from nonplacental tissues in association with the evolution of a novel promoter. Moreover, Mbnl3 has undergone accelerated protein sequence evolution leading to changes in its RNA-binding specificities and cellular localization. While Mbnl2 and Mbnl3 share partially redundant roles in regulating alternative splicing, polyadenylation site usage and, in turn, placenta maturation, Mbnl3 has also acquired novel biological functions. Specifically, Mbnl3 knockout (M3KO) alone results in increased placental growth associated with higher Myc expression. Furthermore, Mbnl3 loss increases fetal resource allocation during limiting conditions, suggesting that location of Mbnl3 on the X chromosome has led to its role in limiting placental growth, favoring the maternal side of the parental genetic conflict. The research has been funded by the Spanish Ministerio de Ciencia (BFU2017-89201-P and PID2020-115040GB-I00 to M.I.), the ‘Centro de Excelencia Severo Ochoa 2013-2017’(SEV-2012-0208 to the Centre for Genomic Regulation), and the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme (FP7/2007-2013) under REA grant agreement n° 608959 to T.S.

Age-related clonal hematopoiesis (ARCH) is characterized by age-associated accumulation of somatic mutations in hematopoietic stem cells (HSCs) or their pluripotent descendants. HSCs harboring driver mutations will be positively selected and cells carrying these mutations will rise in frequency. While ARCH is a known risk factor for blood malignancies, such as Acute Myeloid Leukemia (AML), why some people who harbor ARCH driver mutations do not progress to AML remains unclear. Here, we model the interaction of positive and negative selection in deeply sequenced blood samples from individuals who subsequently progressed to AML, compared to healthy controls, using deep learning and population genetics. Our modeling allows us to discriminate amongst evolutionary classes with high accuracy and captures signatures of purifying selection in most individuals. Purifying selection, acting on benign or mildly damaging passenger mutations, appears to play a critical role in preventing disease-predisposing clones from rising to dominance and is associated with longer disease-free survival. Through exploring a range of evolutionary models, we show how different classes of selection shape clonal dynamics and health outcomes thus enabling us to better identify individuals at a high risk of malignancy., Age-related clonal hematopoiesis is associated with risk for diseases like acute myeloid leukemia (AML), yet it is unclear why some individuals do not progress despite having AML driver mutations. Here, the authors use deep learning and population genetics models to investigate how the interplay of positive and negative selection influences AML progression.

The genetic architecture of blood is highly complex with germline polymorphisms, somatic point mutations, and larger chromosomal alterations playing a role in shaping the fitness of the immune cells. Many advances have been made in understanding how the age associated acquisition of point mutations and somatic structural variants (SSVs) in blood, termed Age-Related Clonal Hematopoiesis (ARCH), predispose individuals to hematological cancer or cardiovascular disease. Yet, ARCH is commonly observed in healthy individuals and our ability to predict who is at risk of progressing to disease remains limited. A previous study which integrated deep learning and population genetics methods to evaluate the complex interplay of selection on point mutations in deeply sequenced blood samples highlighted the role that negative selection plays in prevention progression to hematological cancer. Here, we interrogate the mutational and selective pressures within blood to better understand how the full spectrum of somatic changes within an individual impact clonal fitness and disease outcomes. We evaluate how selection shapes the prevalence of large somatic structural variation in blood sampled from 15,910 individuals across over 20 different genetic ancestries, including the Thousand Genomes Study and the Canadian Partnership for Tomorrow’s Health study. Using dense genotyping arrays, we capture SSVs among 20 populations and find that ARCH attributed to somatic structural variation is three times as high as previously reported with up to 14% of individuals harbouring at least one large SSV in their blood. We estimate the rate at which SSVs accrue in blood cells and find that selection impacts the size and frequency of SSVs within individual blood populations. To determine the functional impact of clonal mutations on molecular phenotypes, we investigate the relationship between structural variation and the transcriptome. We show that gains, losses and copy number neutral variants impact gene expression distinctly, with stabilizing selection shaping the penetrance of copy number alterations in gene expression. Our work shows how different classes of selection shape clonal dynamics in blood thus enabling us to better understand why certain individuals are at a high risk of malignancy. Citation Format: Kimberly Skead, David Soave, Marie-Julie Fave, Vanessa Bruat, Quaid Morris, Philip Awadalla. A multi-omic perspective of how selection shapes blood cancer risk phenotypes in aging populations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2228.

Large chromosomal alterations are common in cancer and often show preferential gain or loss across many cancer types indicating their selective advantage. Triple negative breast cancer (TNBC) exhibits complex mutational spectrum without common oncogenic drivers yet displays consistent loss of large chromosomal regions. Here, we characterize selection pressures that maintain a recurrently deleted region of chromosome 4p in TNBC. We used single cell and bulk WGS phylogenetic analysis of TNCB PT/PDX panel to show that chr4p deletion is an early event in tumor evolution. We used scRNAseq gene expression and inferred copy number analysis to show that chr4p loss is associated with a proliferative state. This finding was confirmed by a combination of RNA in situ hybridization and immunofluorescence. We then tested the dosage sensitivity of genes residing within this region by individual and dual overexpression in TNBC PDX-derived cell lines and control normal cell line by assessing their effect on cell proliferation. The overexpression of genes within chr4p elicited a strong cell proliferation defect in cancer but not normal cell line models. We also characterized an unknown gene within chr4p region as a novel member of the STRIPAK complex. Genome-wide pooled ORFeome library screens identified a global pattern of background-specific dosage sensitive regions. Our study shows that large chromosomal deletions are maintained due to evolutionary early genetic network rewiring rendering multiple genes within such regions to be dosage sensitive. Ultimately, this work enhances our understanding of genetic events that modulate TNBC. Citation Format: Elena Kuzmin, Jean Monlong, Mathieu Bourgey, Jarry Barber, Tom Lesluyes, Toby Baker, Genevieve Morin, Dongmei Zou, Michael Schwartz, Yang Yang, Alain Pacis, Constanza Martinez, Hellen Kuasne, Anne-Marie Fortier, Rui Li, Claudia Kleinman, Sidong Huang, Peter van Loo, Quaid Morris, Jiannis Ragoussis, Guillaume Bourque, Morag Park. Evolution of large copy number variants in breast cancer through genetic network rewiring [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 44.

The DeepTumour algorithm predicts the tissue of origin of a tumor based on the pattern of passenger mutations identified by whole genome sequencing. "Passengers" are incidental mutations that accrue in the genome over time due to random mutational processes, and are functionally distinct from the "driver" mutations that are responsible for the cancer's malignant behavior. In adult cancers, passenger mutations typically outnumber drivers by a hundred or thousand-fold; critically, the vast majority of passengers arise in the normal cell lineage that precedes the malignant transformation event and hence reflects mutational processes existing in the cancer's precursor cell and its ancestors. Passenger mutations are not uniformly distributed across the genome, but are concentrated in areas of the genome that have a locally high mutation rate. Mutation rates are highest at places in the genome where chromatin is tightly packed and less accessible to the DNA repair machinery. Each distinct cell type has a different pattern of chromatin packing due to epigenetic modifications. DeepTumour takes advantage of this to infer the chromatin state in the cell of origin from the distribution of passenger mutations in the tumor. Another characteristic of passenger mutations is that the probability of a particular type of mutation occurring (e.g. replacement of C by T) depends on the mutational processes that were active in the cell of origin and its ancestors. Because certain cancers are associated with distinct mutational exposures (e.g. lung cancer and smoking), DeepTumour uses the tumor's distribution of passenger mutation type as well as position on the genome. The DeepTumour algorithm itself is a fully connected, feed-forward neural network which we trained using 28 cohorts representing different tumor types from the Pan-Cancer Analysis of Whole Genomes project. When applied to independent sets of tumors, the algorithm is able to achieve an overall accuracy of 88% on primary tumors and 83% on metastatic tumors for distinguishing the 28 cancer types. Furthermore, DeepTumour provides estimates of the models uncertainty, allowing it to automatically detect rare cancer samples with an accuracy of 93%. The DeepTumour algorithm is now available as a fast, convenient and secure web-based service at https://deeptumour.oicr.on.ca. It accepts uploads of VCF files containing somatic mutations from tumor whole genome sequencing, and returns a ranked list of tumor type matches and their relative probabilities. It can be used to provide leads when evaluating a cancer of uncertain primary, to assist in resolving diagnostic ambiguities, and as a research tool for understanding tumors of intermediate histology. Citation Format: Lincoln David Stein, Wei Jiao, Gurnit Atwal, Quaid Morris. DeepTumour: Identify tumor origin from whole genome sequences [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5046.

Clinical machine learning models experience significantly degraded performance in datasets not seen during training, e.g., new hospitals or populations. Recent developments in domain generalization offer a promising solution to this problem by creating models that learn invariances across environments. In this work, we benchmark the performance of eight domain generalization methods on multi-site clinical time series and medical imaging data. We introduce a framework to induce synthetic but realistic domain shifts and sampling bias to stress-test these methods over existing non-healthcare benchmarks. We find that current domain generalization methods do not consistently achieve significant gains in out-of-distribution performance over empirical risk minimization on real-world medical imaging data, in line with prior work on general imaging datasets. However, a subset of realistic induced-shift scenarios in clinical time series data do exhibit limited performance gains. We characterize these scenarios in detail, and recommend best practices for domain generalization in the clinical setting., Published at ACM CHIL 2021

Subclonal reconstruction from bulk tumor DNA sequencing has become a pillar of cancer evolution studies, providing insight into the clonality and relative ordering of mutations and mutational processes. We provide an outline of the complex computational approaches used for subclonal reconstruction from single and multiple tumor samples. We identify the underlying assumptions and uncertainties in each step, and suggest best practices for analysis and quality assessment. This guide provides a pragmatic resource for the growing user community of subclonal reconstruction methods.

Juvenile idiopathic arthritis (JIA) encompasses a heterogeneous group of inflammatory disorders linked by chronic joint inflammation [1]. Stratifying patients in a logical manner would improve the ...

1 Abstract Cancers are composed of genetically distinct subpopulations of malignant cells. By sequencing DNA from cancer tissue samples, we can characterize the somatic mutations specific to each population and build clone trees describing the evolutionary ancestry of populations relative to one another. These trees reveal critical points in disease development and inform treatment. Pairtree is a new method for constructing clone trees using DNA sequencing data from one or more bulk samples of an individual cancer. It uses Bayesian inference to compute posterior distributions over the evolutionary relationships between every pair of identified subpopulations, then uses these distributions in a Markov Chain Monte Carlo algorithm to perform efficient inference of the posterior distribution over clone trees. Unlike existing methods, Pairtree can perform clone tree reconstructions using as many as 100 samples per cancer that reveal 30 or more cell subpopulations. On simulated data, Pairtree is the only method whose performance reliably improves when provided with additional bulk samples from a cancer. This suggests a shortcoming of existing methods, as more samples provide more information, and should always make clone tree reconstruction easier. On 14 B-progenitor acute lymphoblastic leukemias with up to 90 samples from each cancer, Pairtree was the only method that could reproduce or improve upon expert-derived clone tree reconstructions. By scaling to more challenging problems, Pairtree supports new biomedical research applications that can improve our understanding of the natural history of cancer, as well as better illustrate the interplay between cancer, host, and therapeutic interventions. The Pairtree method, along with an interactive visual interface for exploring the clone tree posterior, is available at https://github.com/morrislab/pairtree.

Tumors contain multiple subpopulations of genetically distinct cancer cells. Reconstructing their evolutionary history can improve our understanding of how cancers develop and respond to treatment. Subclonal reconstruction methods cluster mutations into groups that co-occur within the same subpopulations, estimate the frequency of cells belonging to each subpopulation, and infer the ancestral relationships among the subpopulations by constructing a clone tree. However, often multiple clone trees are consistent with the data and current methods do not efficiently capture this uncertainty; nor can these methods scale to clone trees with a large number of subclonal populations.Here, we formalize the notion of a partial clone tree that defines a subset of the pairwise ancestral relationships in a clone tree, thereby implicitly representing the set of all clone trees that have these defined pairwise relationships. Also, we introduce a special partial clone tree, the Maximally-Constrained Ancestral Reconstruction (MAR), which summarizes all clone trees fitting the input data equally well. Finally, we extend commonly used clone tree validity conditions to apply to partial clone trees and describe SubMARine, a polynomial-time algorithm producing the subMAR, which approximates the MAR and guarantees that its defined relationships are a subset of those present in the MAR. We also extend SubMARine to work with subclonal copy number aberrations and define equivalence constraints for this purpose. In contrast with other clone tree reconstruction methods, SubMARine runs in time and space that scales polynomially in the number of subclones.We show through extensive simulation and a large lung cancer dataset that the subMAR equals the MAR in > 99.9% of cases where only a single clone tree exists and that it is a perfect match to the MAR in most of the other cases. Notably, SubMARine runs in less than 70 seconds on a single thread with less than one Gb of memory on all datasets presented in this paper, including ones with 50 nodes in a clone tree.The freely-available open-source code implementing SubMARine can be downloaded at https://github.com/morrislab/submarine.Author summaryCancer cells accumulate mutations over time and consist of genetically distinct subpopulations. Their evolutionary history (as represented by tumor phylogenies) can be inferred from bulk cancer genome sequencing data. Current tumor phylogeny reconstruction methods have two main issues: they are slow, and they do not efficiently represent uncertainty in the reconstruction.To address these issues, we developed SubMARine, a fast algorithm that summarizes all valid phylogenies in an intuitive format. SubMARine solved all reconstruction problems in this manuscript in less than 70 seconds, orders of magnitude faster than other methods. These reconstruction problems included those with up to 50 subclones; problems that are too large for other algorithms to even attempt. SubMARine achieves these result because, unlike other algorithms, it performs its reconstruction by identifying an upper-bound on the solution set of trees. In the vast majority of cases, this upper bound is tight: when only a single solution exists, SubMARine converges to it > 99.9% of the time; when multiple solutions exist, our algorithm correctly recovers the uncertain relationships in more than 80% of cases.In addition to solving these two major challenges, we introduce some useful new concepts for and open research problems in the field of tumor phylogeny reconstruction. Specifically, we formalize the concept of a partial clone tree which provides a set of constraints on the solution set of clone trees; and provide a complete set of conditions under which a partial clone tree is valid. These conditions guarantee that all trees in the solution set satisfy the constraints implied by the partial clone tree.

Intra-tumor heterogeneity (ITH) is a mechanism of therapeutic resistance and therefore an important clinical challenge. However, the extent, origin and drivers of ITH across cancer types are poorly understood. To address this question, we extensively characterize ITH across whole-genome sequences of 2,658 cancer samples, spanning 38 cancer types. Nearly all informative samples (95.1%) contain evidence of distinct subclonal expansions, with frequent branching relationships between subclones. We observe positive selection of subclonal driver mutations across most cancer types, and identify cancer type specific subclonal patterns of driver gene mutations, fusions, structural variants and copy-number alterations, as well as dynamic changes in mutational processes between subclonal expansions. Our results underline the importance of ITH and its drivers in tumor evolution, and provide an unprecedented pan-cancer resource of comprehensively annotated subclonal events from whole-genome sequencing data.

The activity of mutational processes differs across the genome, and is influenced by chromatin state and spatial genome organization. At the scale of one megabase-pair (Mb), regional mutation density correlate strongly with chromatin features and mutation density at this scale can be used to accurately identify cancer type. Here, we explore the relationship between genomic region and mutation rate by developing an information theory driven, dynamic programming algorithm for dividing the genome into regions with differing relative mutation rates between cancer types. Our algorithm improves mutual information when compared to the naive approach, effectively reducing the average number of mutations required to identify cancer type. Our approach provides an efficient method for associating regional mutation density with mutation labels, and has future applications in exploring the role of somatic mutations in a number of diseases.

Identification and subsequent intervention of patients at risk of becoming High Cost Users (HCUs) presents the opportunity to improve outcomes while also providing significant savings for the healthcare system. In this paper, the 2016 HCU status of patients was predicted using free-form text data from the 2015 cumulative patient profiles within the electronic medical records of family care practices in Ontario. These unstructured notes make substantial use of domain-specific spellings and abbreviations; we show that word embeddings derived from the same context provide more informative features than pre-trained ones based on Wikipedia, MIMIC, and Pubmed. We further demonstrate that a model using features derived from aggregated word embeddings (EmbEncode) provides a significant performance improvement over the bag-of-words representation (82.48±0.35% versus 81.85±0.36% held-out AUROC, p = 3.2 × 10-4), using far fewer input features (5,492 versus 214,750) and fewer non-zero coefficients (1,177 versus 4,284). The future HCUs of greatest interest are the transitional ones who are not already HCUs, because they provide the greatest scope for interventions. Predicting these new HCU is challenging because most HCUs recur. We show that removing recurrent HCUs from the training set improves the ability of EmbEncode to predict new HCUs, while only slightly decreasing its ability to predict recurrent ones.

Mutations of EXOSC3 have been linked to the rare neurological disorder known as Pontocerebellar Hypoplasia type 1B (PCH1B). EXOSC3 is one of three putative RNA-binding structural cap proteins that guide RNA into the RNA exosome, the cellular machinery that degrades RNA. Using RNAcompete, we identified a G-rich RNA motif binding to EXOSC3. Surface plasmon resonance (SPR) and microscale thermophoresis (MST) indicated an affinity in the low micromolar range of EXOSC3 for long and short G-rich RNA sequences. Although several PCH1B-causing mutations in EXOSC3 did not engage a specific RNA motif as shown by RNAcompete, they exhibited lower binding affinity to G-rich RNA as demonstrated by MST. To test the hypothesis that modification of the RNA-protein interface in EXOSC3 mutants may be phenocopied by small molecules, we performed an in-silico screen of 50 000 small molecules and used enzyme-linked immunosorbant assays (ELISAs) and MST to assess the ability of the molecules to inhibit RNA-binding by EXOSC3. We identified a small molecule, EXOSC3-RNA disrupting (ERD) compound 3 (ERD03), which ( i) bound specifically to EXOSC3 in saturation transfer difference nuclear magnetic resonance (STD-NMR), ( ii) disrupted the EXOSC3-RNA interaction in a concentration-dependent manner, and ( iii) produced a PCH1B-like phenotype with a 50% reduction in the cerebellum and an abnormally curved spine in zebrafish embryos. This compound also induced modification of zebrafish RNA expression levels similar to that observed with a morpholino against EXOSC3. To our knowledge, this is the first example of a small molecule obtained by rational design that models the abnormal developmental effects of a neurodegenerative disease in a whole organism.

Highlights • Oral cancer is highly stratified by tumor as opposed to host characteristics. • Host peripheral blood data can impact prognosis in oral cancer. • Autoimmune diseases are macroscopic reflection of host immune dysregulation. • A score was created known as AI score to depict host status. • AI score was useful as a prognostic tool in oral cancer and should be more widely investigated., Background Recent literature has highlighted the role of the host in prognosis in oral squamous cell carcinoma (OSCC). Autoimmune (AI) disease represents a macroscopic depiction of host status. The goal of this study was to predict an AI “status” and to analyze the utility of this “status” as a prognostic indicator in OSCC. Methods From a departmental database of OSCC patients (n = 1377), 125 patients with an AI disorder were identified. PBL values were obtained and standardized for analysis. A LASSO regression model was used to determine the best predictors of AI status and an AI score was developed. The score was then analyzed across various survival endpoints. Results When AI score was divided into a binary variable, patients in the highest quartile had a significantly worse overall survival (OS), local recurrence-free (LRFP) and distant recurrence-free probability (DRFP). Survival curves showed significant differences for OS, DSS, LRFP, and DRFP. Conclusions AI diseases are immune dysregulations that could play a role in prognosis. Therefore, development of an AI score is necessary to depict host status in a ubiquitous manner. AI score as a binary variable may be more utilitarian in a clinical setting, compared to the continuous score. This novel tool needs validation and integration into more tumor and host characteristics. This investigation showed utility of such a score, similar to PBL data in OSCC prognosis. Future studies should incorporate other relevant variables known to affect outcome and implement a more comprehensive predictive model.

We recognize that many cancers are highly complex mixtures of sub-populations of cells, which are also influenced by their microenvironment. This heterogeneity explains in part, why the majority of current therapeutic approaches for cancer work best when multiple agents are combined. Therefore, in an era of targeted therapeutics it becomes critical to understand the complexity of tumors both at diagnosis and over the course of therapy, including measures of heterogeneity. Here we present genomic, transcriptomic and in situ proteomic profiling of a cohort of breast cancer (BCa) lumpectomies to reveal the extent of heterogeneity in pathologically defined unifocal and multifocal cancers. Integration of molecular profiling, phylogenetic analyses and radiomics has the potential to significantly improve BCa clinical management and stratification to targeted therapies that are already available in the clinic. In this study, lumpectomies with imaging data were processed as whole mounts with serial blocks reviewed. Tissue cores were taken from at least three different regions through the lumpectomy for nucleic acid extraction and tissue microarray (TMA) construction, focusing on morphologic/histological differences in addition to the spatial orientation of the sampled region within the lumpectomy. Targeted sequencing using the Oncomine Comprehensive Assay v3 (OCAv3); transcriptional profiling using the NanoString Breast Cancer 360 Panel; in situ profiling by multiplex fluorescence immunohistochemistry (MxIF) and Digital Spatial Profiling (see abstract Spears et al) were performed. From this cohort of 60 patients, we present a subset of patients demonstrating integration of the molecular profiling to reveal the phylogenetic relationship between the multiple samplings and the impact on clinical decision making in BCa. Briefly, we identified differences in the molecular subtypes between the different sample regions from the same unifocal cancer as well as differences in the predicted responses to anti-PDL1 therapy by transcriptional profiling; while targeted sequencing of driver mutations suggested the likelihood of an ancestral tumor cell giving rise to the lesions in pathologically defined multifocal cancers. However it was evident that genes and pathways found to be aberrant in these different lesions from the same cancer could impact the response to standard BCa treatment, or the selection of targeted therapies. In situ proteomics demonstrated differences in the expression of standard BCa markers ER, PgR, HER2 and Ki67 in addition to immune markers in the tumor and tumor microenvironment. While there are clinically validated transcriptional risk test available for BCa, we have demonstrated that transcriptomics or genomics alone is insufficient for a rational stratification of patients to currently available targeted therapies; therefore supporting the need for an integrative approach. Citation Format: Jane Bayani, Quang M. Trinh, Mary Anne Quintayo, Cheryl Crozier, Megan Hopkins, Jingping Qiao, Alison Cheung, James Mainprize, Quaid Morris, Melanie Spears, Martin Yaffe, Lincoln Stein, John M. Bartlett. Tumour spatial heterogeneity in breast cancer and the impact on clinical management [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3131.

RNA-binding proteins recognize RNA sequences and structures, but there is currently no systematic and accurate method to derive large (>12base) motifs de novo that reflect a combination of intrinsic preference to both sequence and structure. To address this absence, we introduce RNAcompete-S, which couples a single-step competitive binding reaction with an excess of random RNA 40-mers to a custom computational pipeline for interrogation of the bound RNA sequences and derivation of SSMs (Sequence and Structure Models). RNAcompete-S confirms that HuR, QKI, and SRSF1 prefer binding sites that are single stranded, and recapitulates known 8-10bp sequence and structure preferences for Vts1p and RBMY. We also derive an 18-base long SSM for Drosophila SLBP, which to our knowledge has not been previously determined by selections from pure random sequence, and accurately discriminates human replication-dependent histone mRNAs. Thus, RNAcompete-S enables accurate identification of large, intrinsic sequence-structure specificities with a uniform assay.

RNA-binding proteins (RBPs) participate in diverse cellular processes and have important roles in human development and disease. The human genome, and that of many other eukaryotes, encodes hundreds of RBPs that contain canonical sequence-specific RNA-binding domains (RBDs) as well as numerous other unconventional RNA binding proteins (ucRBPs). ucRBPs physically associate with RNA but lack common RBDs. The degree to which these proteins bind RNA, in a sequence specific manner, is unknown. Here, we provide a detailed description of both the laboratory and data processing methods for RNAcompete, a method we have previously used to analyze the RNA binding preferences of hundreds of RBD-containing RBPs, from diverse eukaryotes. We also determine the RNA-binding preferences for two human ucRBPs, NUDT21 and CNBP, and use this analysis to exemplify the RNAcompete pipeline. The results of our RNAcompete experiments are consistent with independent RNA-binding data for these proteins and demonstrate the utility of RNAcompete for analyzing the growing repertoire of ucRBPs.

Identification and subsequent intervention of patients at risk of becoming High Cost Users (HCUs) presents the opportunity to improve outcomes while also providing significant savings for the healthcare system. In this paper, the 2016 HCU status of patients was predicted using free-form text data from the 2015 cumulative patient profiles within the electronic medical records of family care practices in Ontario. These unstructured notes make substantial use of domain-specific spellings and abbreviations; we show that word embeddings derived from the same context provide more informative features than pre-trained ones based on Wikipedia, MIMIC, and Pubmed. We further demonstrate that a model using features derived from aggregated word embeddings (EmbEncode) provides a significant performance improvement over the bag-of-words representation (82.48±0.35% versus 81.85±0.36% held-out AUROC, p = 3.2 × 10-4), using far fewer input features (5,492 versus 214,750) and fewer non-zero coefficients (1,177 versus 4,284). The future HCUs of greatest interest are the transitional ones who are not already HCUs, because they provide the greatest scope for interventions. Predicting these new HCU is challenging because most HCUs recur. We show that removing recurrent HCUs from the training set improves the ability of EmbEncode to predict new HCUs, while only slightly decreasing its ability to predict recurrent ones.

Disease recurrence causes significant mortality in B-progenitor acute lymphoblastic leukemia (B-ALL). Genomic analysis of matched diagnosis and relapse samples shows relapse often arising from minor diagnosis subclones. However, why therapy eradicates some subclones while others survive and progress to relapse remains obscure. Elucidation of mechanisms underlying these differing fates requires functional analysis of isolated subclones. Here, large-scale limiting dilution xenografting of diagnosis and relapse samples, combined with targeted sequencing, identified and isolated minor diagnosis subclones that initiate an evolutionary trajectory toward relapse [termed diagnosis Relapse Initiating clones (dRI)]. Compared with other diagnosis subclones, dRIs were drug-tolerant with distinct engraftment and metabolic properties. Transcriptionally, dRIs displayed enrichment for chromatin remodeling, mitochondrial metabolism, proteostasis programs, and an increase in stemness pathways. The isolation and characterization of dRI subclones reveals new avenues for eradicating dRI cells by targeting their distinct metabolic and transcriptional pathways before further evolution renders them fully therapy-resistant. Significance: Isolation and characterization of subclones from diagnosis samples of patients with B-ALL who relapsed showed that relapse-fated subclones had increased drug tolerance and distinct metabolic and survival transcriptional programs compared with other diagnosis subclones. This study provides strategies to identify and target clinically relevant subclones before further evolution toward relapse. See related video: https://vimeo.com/442838617 See related article by E. Waanders et al .

Regulation of translation during early human development is poorly understood and its dysregulation is associated with Rett syndrome (RTT). To discover shifts in mRNA ribosomal engagement (RE) that occur during human neurodevelopment and disease, we used parallel TRAP-Seq and RNA-Seq on control and RTT human induced pluripotent stem cells, neural progenitor cells and cortical neurons. We found 30% of transcribed genes were translationally regulated including key gene sets (neurodevelopment, transcription and translation factors, glycolysis). Neurons more efficiently translated global mRNAs, had longer 3`UTRs and RE correlated with RNA-binding protein motif composition. In contrast, RTT neurons had reduced global translation, compromised mTOR signaling and over 2100 genes were translationally dysregulated. NEDD4L E3-ubiquitin ligase was most translationally impaired, ubiquitinated protein levels were reduced, and protein targets accumulated in RTT neurons. Our findings reveal a dynamic translatome in neurodevelopment that is disturbed in RTT, and provide novel insight into altered proteostasis that may have therapeutic implications.

Cancer cells enter a reversible drug-tolerant persister (DTP) state to evade death from chemotherapy and targeted agents. It is increasingly appreciated that DTPs are important drivers of therapy failure and tumor relapse. We combined cellular barcoding and mathematical modeling in patient-derived colorectal cancer models to identify and characterize DTPs in response to chemotherapy. Barcode analysis revealed no loss of clonal complexity of tumors that entered the DTP state and recurred following treatment cessation. Our data fit a mathematical model where all cancer cells, and not a small subpopulation, possess an equipotent capacity to become DTPs. Mechanistically, we determined that DTPs display remarkable transcriptional and functional similarities to diapause, a reversible state of suspended embryonic development triggered by unfavorable environmental conditions. Our study provides insight into how cancer cells use a developmentally conserved mechanism to drive the DTP state, pointing to novel therapeutic opportunities to target DTPs.

The age-associated accumulation of somatic mutations and large-scale structural variants (SVs) in the early hematopoietic hierarchy have been linked to premalignant stages for cancer and cardiovascular disease (CVD). However, only a small proportion of individuals harboring these mutations progress to disease, and mechanisms driving the transformation to malignancy remains unclear. Hematopoietic evolution, and cancer evolution more broadly, has largely been studied through a lens of adaptive evolution and the contribution of functionally neutral or mildly damaging mutations to early disease-associated clonal expansions has not been well characterised despite comprising the majority of the mutational burden in healthy or tumoural tissues. Through combining deep learning with population genetics, we interrogate the hematopoietic system to capture signatures of selection acting in healthy and pre-cancerous blood populations. Here, we leverage high-coverage sequencing data from healthy and pre-cancerous individuals from the European Prospective Investigation into Cancer and Nutrition Study (n=477) and dense genotyping from the Canadian Partnership for Tomorrow's Health (n=5,000) to show that blood rejects the paradigm of strictly adaptive or neutral evolution and is subject to pervasive negative selection. We observe clear age associations across hematopoietic populations and the dominant class of selection driving evolutionary dynamics acting at an individual level. We find that both the location and ratio of passenger to driver mutations are critical in determining if positive selection acting on driver mutations is able to overwhelm regulated hematopoiesis and allow clones harbouring disease-predisposing mutations to rise to dominance. Certain genes are enriched for passenger mutations in healthy individuals fitting purifying models of evolution, suggesting that the presence of passenger mutations in a subset of genes might confer a protective role against disease-predisposing clonal expansions. Finally, we find that the density of gene disruption events with known pathogenic associations in somatic SVs impacts the frequency at which the SV segregates in the population with variants displaying higher gene disruption density segregating at lower frequencies. Understanding how blood evolves towards malignancy will allow us to capture cancer in its earliest stages and identify events initiating departures from healthy blood evolution. Further, as the majority of mutations are passengers, studying their contribution to tumorigenesis, will unveil novel therapeutic targets thus enabling us to better understand patterns of clonal evolution in order to diagnose and treat disease in its infancy. Disclosures Dick: Bristol-Myers Squibb/Celgene: Research Funding.

Many cancers are highly complex mixtures of many sub-populations of cells, also influenced by their microenvironment. This heterogeneity explains, in part, why the majority of current therapeutic approaches for cancer work best when multiple agents are combined. Therefore, in an era of targeted therapeutics it becomes critical to understand the complexity of tumours both at diagnosis and over the course of therapy, including measures of heterogeneity. Here we present genomic, transcriptomic and in situ proteomic profiling of a cohort of breast cancer (BCa) lumpectomies with associated imaging data to reveal the extent of heterogeneity in pathologically defined unifocal and multifocal cancers. Integration of molecular profiling, phylogenetic analyses and radiomics has the potential to significantly improve BCa clinical management and stratification to targeted therapies that are already available in the clinic. In this study, lumpectomies were processed as whole mounts with serial blocks reviewed. Tissue cores were taken from at least three different regions throughout the lumpectomy for nucleic acid extraction and tissue microarray (TMA) construction, focusing on morphologic/histological differences in addition to the spatial orientation of the sampled region within the lumpectomy. Targeted sequencing using the Oncomine Comprehensive Assay v3 (OCAv3); transcriptional profiling using the NanoString Breast Cancer 360 Panel; and in situ profiling by multiplex fluorescence immunohistochemistry (MxIF) performed (see abstract Cheung et al). From this cohort of 60 patients, we present a subset of patients demonstrating integration of the molecular profiling to reveal the phylogenetic relationship between the multiple samplings and the potential impact on clinical decision making in BCa. We identified differences in the molecular subtypes between the different sample regions from the same unifocal cancer as well as differences in the predicted responses to anti-PDL1 therapy by transcriptional profiling. Targeted sequencing of driver mutations suggested the likelihood of an ancestral tumour cell giving rise to the lesions in pathologically defined multifocal cancers; however it was evident that genes and pathways found to be aberrant in these different lesions from the same cancer could impact the response to standard BCa treatment, or the selection of targeted therapies. In situ proteomics demonstrated differences in the expression of standard BCa markers ER, PgR, HER2 and Ki67, in addition to immune markers in the tumour and its microenvironment. While there are clinically validated transcriptional risk tests available for BCa, we have demonstrated that transcriptomics or genomics alone is insufficient for a rational stratification of patients to currently available targeted therapies, therefore, supporting the need for an integrative approach. Citation Format: Jane Bayani, Quang M. Trinh, Mary Anne Quintayo, Cheryl Crozier, Megan Hopkins, Jingping Qiao, Alison Cheung, James G Mainprize, Quaid Morris, Melanie Spears, Martin J. Yaffe, Lincoln D. Stein, John M.S. Bartlett. Revealing tumour spatial heterogeneity in breast cancer and the impact on clinical management [abstract]. In: Proceedings of the AACR Virtual Special Conference on Tumor Heterogeneity: From Single Cells to Clinical Impact; 2020 Sep 17-18. Philadelphia (PA): AACR; Cancer Res 2020;80(21 Suppl):Abstract nr PO-002.