Search

Malignant peripheral nerve sheath tumor (MPNST), an aggressive soft-tissue sarcoma, occurs in people with neurofibromatosis type 1 (NF1) and sporadically. Whole-genome and multiregional exome sequencing, transcriptomic, and methylation profiling of 95 tumor samples revealed the order of genomic events in tumor evolution. Following biallelic inactivation of NF1, loss of CDKN2A or TP53 with or without inactivation of polycomb repressive complex 2 (PRC2) leads to extensive somatic copy-number aberrations (SCNA). Distinct pathways of tumor evolution are associated with inactivation of PRC2 genes and H3K27 trimethylation (H3K27me3) status. Tumors with H3K27me3 loss evolve through extensive chromosomal losses followed by whole-genome doubling and chromosome 8 amplification, and show lower levels of immune cell infiltration. Retention of H3K27me3 leads to extensive genomic instability, but an immune cell-rich phenotype. Specific SCNAs detected in both tumor samples and cell-free DNA (cfDNA) act as a surrogate for H3K27me3 loss and immune infiltration, and predict prognosis.Significance:MPNST is the most common cause of death and morbidity for individuals with NF1, a relatively common tumor predisposition syndrome. Our results suggest that somatic copy-number and methylation profiling of tumor or cfDNA could serve as a biomarker for early diagnosis and to stratify patients into prognostic and treatment-related subgroups.This article is highlighted in the In This Issue feature, p. 517

Supplementary Table 1 includes the metadata on all subjects, including basic demographics and medical history. Supplementary Table 2 provides statistics on chrom 8 gene expression among tumors with H3K27me3 loss. Note: no changes have been made to this file since our last resubmission.

Supplementary figures provide additional details on the genomic landscape of MPNST and interesting cases within the cohort.

Malignant peripheral nerve sheath tumor (MPNST), an aggressive soft-tissue sarcoma, occurs in people with neurofibromatosis type 1 (NF1) and sporadically. Whole-genome and multiregional exome sequencing, transcriptomic, and methylation profiling of 95 tumor samples revealed the order of genomic events in tumor evolution. Following biallelic inactivation of NF1, loss of CDKN2A or TP53 with or without inactivation of polycomb repressive complex 2 (PRC2) leads to extensive somatic copy-number aberrations (SCNA). Distinct pathways of tumor evolution are associated with inactivation of PRC2 genes and H3K27 trimethylation (H3K27me3) status. Tumors with H3K27me3 loss evolve through extensive chromosomal losses followed by whole-genome doubling and chromosome 8 amplification, and show lower levels of immune cell infiltration. Retention of H3K27me3 leads to extensive genomic instability, but an immune cell-rich phenotype. Specific SCNAs detected in both tumor samples and cell-free DNA (cfDNA) act as a surrogate for H3K27me3 loss and immune infiltration, and predict prognosis.Significance:MPNST is the most common cause of death and morbidity for individuals with NF1, a relatively common tumor predisposition syndrome. Our results suggest that somatic copy-number and methylation profiling of tumor or cfDNA could serve as a biomarker for early diagnosis and to stratify patients into prognostic and treatment-related subgroups.This article is highlighted in the In This Issue feature, p. 517

Whole-genome sequencing studies of human tumours have revealed that complex forms of structural variation, collectively known as complex genome rearrangements (CGR), are pervasive across diverse cancer types. Detection, classification, and mechanistic interpretation of CGR requires the visualization of complex patterns of somatic copy number aberrations (SCNAs) and structural variants (SVs). However, there is a lack of tools specifically designed to facilitate the visualization and study of CGR. We present ReConPlot (REarrangement and COpy Number PLOT), an R package that provides functionalities for the joint visualization of SCNAs and SVs across one or multiple chromosomes. ReConPlot is based on the popular ggplot2 package, thus allowing customization of plots and the generation of publication-quality figures with minimal effort. Overall, ReConPlot facilitates the exploration, interpretation, and reporting of complex genome rearrangement patterns. Code availability: The R package ReConPlot is available at https://github.com/cortes-ciriano-lab/GenomicRearrangementPlot. Detailed documentation and a tutorial with examples are provided with the package.

The JAK2-V617F mutation is the most common cause of myeloproliferative neoplasms. While experiments have shown that this gain-of-function mutation is associated with myeloid blood cell expansion and increased production of white cells, red cells and platelets, the transcriptional consequences of the JAK2-V617F mutation in different cellular compartments of the bone marrow have not yet been fully elucidated. To study the direct effects of JAK2-V617F on bone marrow cells in myeloproliferative neoplasm patients, we performed joint single-cell RNA sequencing and JAK2 genotyping on CD34+ enriched cells from 8 patients with newly diagnosed essential thrombocythemia or polycythemia vera. We found that the JAK2-V617F mutation increases the expression of interferon-response genes (e.g., HLAs) and the leptin receptor in hematopoietic progenitor cells. Furthermore, we sequenced a population of CD34-bone marrow monocytes and found the JAK2 mutation increased expression of intermediate monocyte genes and the fibrocyte-associated surface protein SLAMF7 in these cells.

CA extern

Characterization of somatic mutations at single-cell resolution is essential to study cancer evolution, clonal mosaicism, and cell plasticity. However, detection of mutations in single cells remains technically challenging. Here, we describe SComatic, an algorithm designed for the detection of somatic mutations in single-cell transcriptomic and ATAC-seq data sets without requiring matched bulk or single-cell DNA sequencing data. Using > 1.5M single cells from 383 single-cell RNAseq and single-cell ATAC-seq data sets spanning cancer and non-neoplastic samples, we show that SComatic detects mutations in single cells, even in differentiated cells from polyclonal tissues not amenable to mutation detection using existing methods. In addition, SComatic permits the estimation of mutational burdens and de novo mutational signature analysis at single-cell and cell-type resolution. Notably, using matched exome and single-cell RNAseq data, we show that SComatic achieves a 20 to 40-fold increase in precision as compared to existing algorithms for somatic SNV calling without compromising sensitivity. Overall, SComatic opens the possibility to study somatic mutagenesis at unprecedented scale and resolution using high-throughput single-cell profiling data sets.

Malignant peripheral nerve sheath tumour (MPNST) is an aggressive soft-tissue sarcoma that arises in peripheral nerves. MPNST occurs either sporadically or in people with neurofibromatosis type 1 (NF1), a common cancer predisposition syndrome caused by germline pathogenic variants in NF1. Although MPNST is the most common cause of death and morbidity for individuals with NF1, the molecular underpinnings of MPNST pathogenesis remain unclear. Here, we report the analysis of whole-genome sequencing, multi-regional exome sequencing, transcriptomic and methylation profiling data for 95 MPNSTs and precursor lesions (64 NF1-related; 31 sporadic) from 77 individuals. Early events in tumour evolution include biallelic inactivation of NF1 followed by inactivation of CDKN2A and in some cases also TP53 and polycomb repressive complex 2 (PRC2) genes. Subsequently, both sporadic and NF1-related MPNSTs acquire a high burden of somatic copy number alterations (SCNAs). Our analysis revealed distinct pathways of tumour evolution and immune infiltration associated with inactivation of PRC2 genes and H3K27 trimethylation (H3K27me3) status. Tumours with loss of H3K27me3 evolve through extensive chromosomal losses with retention of chromosome 8 heterozygosity followed by whole genome doubling and chromosome 8 amplification. These tumours show lower levels of immune cell infiltration with low cytotoxic activity and low expression of immune checkpoints. In contrast, tumours with retention of H3K27me3 evolve through extensive genomic instability in the absence of recurrent alterations and exhibit an immune cell-rich phenotype. Specific SCNAs detected in both tumour samples and cell-free DNA (cfDNA) act as a surrogate for loss of H3K27me3 and immune infiltration, and predict prognosis. Our results suggest that SCNA profiling of tumour or cfDNA could serve as a biomarker for early diagnosis and to stratify patients into prognostic and treatment-related subgroups.

Although artificial intelligence (AI) has had a profound impact on areas such as image recognition, comparable advances in drug discovery are rare. This article quantifies the stages of drug discovery in which improvements in the time taken, success rate or affordability will have the most profound overall impact on bringing new drugs to market. Changes in clinical success rates will have the most profound impact on improving success in drug discovery; in other words, the quality of decisions regarding which compound to take forward (and how to conduct clinical trials) are more important than speed or cost. Although current advances in AI focus on how to make a given compound, the question of which compound to make, using clinical efficacy and safety-related end points, has received significantly less attention. As a consequence, current proxy measures and available data cannot fully utilize the potential of AI in drug discovery, in particular when it comes to drug efficacy and safety in vivo. Thus, addressing the questions of which data to generate and which end points to model will be key to improving clinically relevant decision-making in the future.

Microsatellite instability (MSI) is caused by deficient DNA mismatch repair (MMR) and is a ubiquitous feature of cancer. Werner syndrome (WRN) helicase is involved in genome stability and DNA repair. We identified WRN as a synthetic-lethal target in dMMR/MSI cancers and highlighted WRN inhibition as a therapeutic option for dMMR/MSI cancers refractory to available therapies. A previously unappreciated genetic feature of dMMR/MSI cancer cells, DNA (TA)n-dinucleotide repeat expansions, were recently reported to cause vulnerability to WRN depletion. Our mechanistic understanding of TA-dinucleotide repeat expansion biology is limited, and their potential therapeutic implications are unclear. To investigate the landscape of these alterations in cancer, we inferred (TA)-dinucleotide repeat expansions by performing coverage analysis in a collection of hundreds of preclinical cancer models and human tumors (PCAWG) profiled by whole genome sequencing (WGS). We validated our findings in cancer cell lines and organoid cultures by performing long-read WGS. Furthermore, we investigated TA-expansions in single-cell-derived clones from human MSI tumors and cancer organoids. Finally, we inferred TA repeats in laser capture microdissection (LCMB)-derived samples obtained from patients affected by familial cancer predisposition syndromes. Our analysis unveils the landscape of TA-repeats alterations in a large collection of tumors and preclinical models, informing on the level of inter-tumor heterogeneity and their association with variable levels of WRN dependency. In addition, we investigated intra-patient tumor heterogeneity of TA-repeats length both within clonal organoids expanded from normal and neoplastic colorectal stem cells, and within different subclones derived from MSI cancer organoids. Furthermore, analysis of non-neoplastic and neoplastic tissues from patients affected by familial cancer predisposition syndromes revealed the pattern of TA-repeats expansions associated with various DNA-repair pathway alterations. Finally, we will discuss the clinical implications of our findings, as TA-repeats heterogeneity may affect sensitivity and resistance to the future generation of WRN inhibitors. Our data provide fresh insights into the inter and intra-tumoral heterogeneity of TA-dinucleotide repeat expansions in human cancers. These data contribute to understanding the role of MMR in cancer and exploiting Werner as a therapeutic target in cancer. Citation Format: Gabriele Picco, Shriram Bhosle, Maria Kalyva, Elena Grassi, Freddy Gibson, Angham Al Saedi, Sara Vieira, Mathijs Sanders, Livio Trusolino, Andrea Bertotti, Isidro Cortes-Ciriano, Mathew Garnett. MSI cancer associated DNA (TA)n-dinucleotide repeat expansions and implications for Werner synthetic lethality. [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 4327.

Graphical abstract, Highlights • Drug discovery data and data from other sources are different in quantity and characteristics. • This article underlines the difference of data from different domains. • In order to fully benefit from algorithms we need to address challenges posed by the data. • Data should not be detached from a hypothesis, its representation, and the method in a data-scarce area., ‘Artificial Intelligence’ (AI) has recently had a profound impact on areas such as image and speech recognition, and this progress has already translated into practical applications. However, in the drug discovery field, such advances remains scarce, and one of the reasons is intrinsic to the data used. In this review, we discuss aspects of, and differences in, data from different domains, namely the image, speech, chemical, and biological domains, the amounts of data available, and how relevant they are to drug discovery. Improvements in the future are needed with respect to our understanding of biological systems, and the subsequent generation of practically relevant data in sufficient quantities, to truly advance the field of AI in drug discovery, to enable the discovery of novel chemistry, with novel modes of action, which shows desirable efficacy and safety in the clinic.

Motivation Quantitative structure–activity relationship (QSAR) methods are increasingly used in assisting the process of preclinical, small molecule drug discovery. Regression models are trained on data consisting of a finite-dimensional representation of molecular structures and their corresponding target-specific activities. These supervised learning models can then be used to predict the activity of previously unmeasured novel compounds. Results This work provides methods that solve three problems in QSAR modelling: (i) a method for comparing the information content between finite-dimensional representations of molecular structures (fingerprints) with respect to the target of interest, (ii) a method that quantifies how the accuracy of the model prediction degrades as a function of the distance between the testing and training data and (iii) a method to adjust for screening dependent selection bias inherent in many training datasets. For example, in the most extreme cases, only compounds which pass an activity-dependent screening threshold are reported. A semi-supervised learning framework combines (ii) and (iii) and can make predictions, which take into account the similarity of the testing compounds to those in the training data and adjust for the reporting selection bias. We illustrate the three methods using publicly available structure–activity data for a large set of compounds reported by GlaxoSmithKline (the Tres Cantos AntiMalarial Set, TCAMS) to inhibit asexual in vitro Plasmodium falciparum growth. Availabilityand implementation https://github.com/owatson/PenalizedPrediction. Supplementary information Supplementary data are available at Bioinformatics online.

A high tumour mutational burden (hypermutation) is observed in some gliomas1–5; however, the mechanisms by which hypermutation develops and whether it predicts the response to immunotherapy are poorly understood. Here we comprehensively analyse the molecular determinants of mutational burden and signatures in 10,294 gliomas. We delineate two main pathways to hypermutation: a de novo pathway associated with constitutional defects in DNA polymerase and mismatch repair (MMR) genes, and a more common post-treatment pathway, associated with acquired resistance driven by MMR defects in chemotherapy-sensitive gliomas that recur after treatment with the chemotherapy drug temozolomide. Experimentally, the mutational signature of post-treatment hypermutated gliomas was recapitulated by temozolomide-induced damage in cells with MMR deficiency. MMR-deficient gliomas were characterized by a lack of prominent T cell infiltrates, extensive intratumoral heterogeneity, poor patient survival and a low rate of response to PD-1 blockade. Moreover, although bulk analyses did not detect microsatellite instability in MMR-deficient gliomas, single-cell whole-genome sequencing analysis of post-treatment hypermutated glioma cells identified microsatellite mutations. These results show that chemotherapy can drive the acquisition of hypermutated populations without promoting a response to PD-1 blockade and supports the diagnostic use of mutational burden and signatures in cancer. Temozolomide therapy seems to lead to mismatch repair deficiency and hypermutation in gliomas, but not to an increase in response to immunotherapy.

Telomere fusions (TFs) can trigger the accumulation of diverse genomic rearrangements and the acquisition of oncogenic alterations leading to malignant transformation and resistance to chemotherapy. Despite their relevance in tumour evolution, our understanding of the patterns and consequences of TFs in human cancer remains limited. Here, we have characterized the rates and spectrum of somatic TFs across >30 cancer types using whole-genome sequencing data. TFs are pervasive in human tumours with rates varying markedly across and within cancer types. In addition to end-to-end fusions, we find novel patterns of TFs that we mechanistically link to the activity of the alternative lengthening of telomeres (ALT) pathway. We show that TFs can be detected in the blood of cancer patients, which enables cancer detection with high specificity and sensitivity even for early-stage tumours and cancer types for which early detection remains a high unmet clinical need, such as pancreatic cancer and brain tumours. Overall, we report a novel genomic footprint that enables characterization of the telomere maintenance mechanism of tumours and liquid biopsy analysis, which has implications for early detection, prognosis, and treatment selection.

DNA mismatch repair deficiency (MMRd) is associated with high tumor mutational burden (TMB), increased T cell infiltration, and remarkable responsiveness to immune checkpoint blockade (ICB) therapy1. Nevertheless, about half of MMRd tumors do not respond to ICB for unclear reasons. While cell line transplant models of MMRd have reinforced the importance of TMB in immune response2,3, critical questions remain regarding the role of immunosurveillance in the evolution of MMRd tumors induced in vivo. Here, we developed autochthonous mouse models of lung and colon cancer with ablation of MMR via in vivo CRISPR/Cas9 targeting. Surprisingly, MMRd in these models did not increase T cell infiltration or response to ICB. Mechanistically, we showed this lack of immunogenicity to be driven by profound intratumoral heterogeneity. Studies in immune deficient animals further demonstrated that immunosurveillance in MMRd tumors has no impact on TMB but shapes the clonal architecture of neoantigens by exacerbating heterogeneity. These results provide important context for understanding immune evasion in cancers with high TMB and have major implications for therapies aimed at increasing TMB.

DNA mismatch repair deficiency (MMRd) in human cancer is associated with high tumor mutational burden (TMB), frameshift mutation-derived neoantigens, increased T cell infiltration, and remarkable responsiveness to immune checkpoint blockade (ICB) therapy. Nevertheless, about half of MMRd tumors do not respond to ICB for unclear reasons. While tumor cell line transplant models of MMRd have reinforced the importance of TMB in immune response, critical questions remain regarding the role of immunosurveillance in the evolution of MMRd tumors induced in vivo. Here, we developed autochthonous mouse models of lung and colon cancer with highly efficient ablation of MMR genes via in vivo CRISPR/Cas9 targeting. Surprisingly, MMRd in these models did not result in increased immunogenicity or response to ICB. Mechanistically, we showed this lack of immunogenicity to be driven by profound intratumoral heterogeneity (ITH). Studies in animals depleted of T cells further demonstrated that immunosurveillance in MMRd tumors has no impact on TMB but shapes the clonal architecture of neoantigens by exacerbating ITH. These results provide important context for understanding immune evasion in cancers with high TMB and have major implications for therapies aimed at increasing TMB.

Dissemination of cancer cells from primary tumors to the brain occurs in many cancer patients, increasing morbidity and death. There is an unmet medical need to develop translational platforms to evaluate therapeutic responses. Toward this goal, we established a library of 23 patient-derived xenografts (PDXs) of brain metastases (BMs) from eight distinct primary tumors. In vivo tumor formation correlates with patients' poor survival. Mouse subcutaneous xenografts develop spontaneous metastases and intracardiac PDXs increase dissemination to the CNS, both models mimicking the dissemination pattern of the donor patient. We test the FDA-approved drugs buparlisib (pan-PI3K inhibitor) and everolimus (mTOR inhibitor) and show their efficacy in treating our models. Finally, we show by RNA sequencing that human BMs and their matched PDXs have similar transcriptional profiles. Overall, these models of BMs recapitulate the biology of human metastatic disease and can be valuable translational platforms for precision medicine.

Affinity fingerprints report the activity of small molecules across a set of assays, and thus permit to gather information about the bioactivities of structurally dissimilar compounds, where models based on chemical structure alone are often limited, and model complex biological endpoints, such as human toxicity and in vitro cancer cell line sensitivity. Here, we propose to model in vitro compound activity using computationally predicted bioactivity profiles as compound descriptors. To this aim, we apply and validate a framework for the calculation of QSAR-derived affinity fingerprints (QAFFP) using a set of 1360 QSAR models generated using Ki, Kd, IC50 and EC50 data from ChEMBL database. QAFFP thus represent a method to encode and relate compounds on the basis of their similarity in bioactivity space. To benchmark the predictive power of QAFFP we assembled IC50 data from ChEMBL database for 18 diverse cancer cell lines widely used in preclinical drug discovery, and 25 diverse protein target data sets. This study complements part 1 where the performance of QAFFP in similarity searching, scaffold hopping, and bioactivity classification is evaluated. Despite being inherently noisy, we show that using QAFFP as descriptors leads to errors in prediction on the test set in the ~ 0.65–0.95 pIC50 units range, which are comparable to the estimated uncertainty of bioactivity data in ChEMBL (0.76–1.00 pIC50 units). We find that the predictive power of QAFFP is slightly worse than that of Morgan2 fingerprints and 1D and 2D physicochemical descriptors, with an effect size in the 0.02–0.08 pIC50 units range. Including QSAR models with low predictive power in the generation of QAFFP does not lead to improved predictive power. Given that the QSAR models we used to compute the QAFFP were selected on the basis of data availability alone, we anticipate better modeling results for QAFFP generated using more diverse and biologically meaningful targets. Data sets and Python code are publicly available at https://github.com/isidroc/QAFFP_regression.

Mutations in BRCA1 and/or BRCA2 (BRCA1/2) are the most common indication of deficiency in the homologous recombination (HR) DNA repair pathway. However, recent genome-wide analyses have shown that the same pattern of mutations found in BRCA1/2-mutant tumors is also present in several other tumors. Here, we present a new computational tool called Signature Multivariate Analysis (SigMA), which can be used to accurately detect the mutational signature associated with HR deficiency from targeted gene panels. Whereas previous methods require whole-genome or whole-exome data, our method detects the HR-deficiency signature even from low mutation counts, by using a likelihood-based measure combined with machine-learning techniques. Cell lines that we identify as HR deficient show a significant response to poly (ADP-ribose) polymerase (PARP) inhibitors; patients with ovarian cancer whom we found to be HR deficient show a significantly longer overall survival with platinum regimens. By enabling panel-based identification of mutational signatures, our method substantially increases the number of patients that may be considered for treatments targeting HR deficiency.

Some cancers originate from a single mutation event in a single cell. For example, blood cancers known as myeloproliferative neoplasms (MPN) are thought to originate through the acquisition of a driver mutation (most commonly JAK2-V617F) in a hematopoietic stem cell (HSC). However, when the mutation first occurs in individual patients and how it impacts the behavior of HSCs in their native context is not known. Here we quantified the impact of the JAK2-V617F mutation on the proliferation dynamics of HSCs and the differentiation trajectories of their progenies in individual MPN patients. We reconstructed the lineage history of individual HSCs obtained from MPN patients using the patterns of spontaneous somatic mutations accrued in their genomes over time. Strikingly, we found that the JAK2-V617F mutation occurred in a single HSC several decades before MPN diagnosis — at age 9±2 years in a 34-year-old patient, and at age 19±3 years in a 63-year-old patient. For each patient, we inferred the number of mutated HSCs over time and computed their fitness. The population of JAK2-mutated HSCs grew exponentially by 63±15% and 44±13% every year in the two patients, respectively. To contrast the differentiation trajectories of the JAK2-mutated HSCs with those of healthy HSCs, we simultaneously measured the full transcriptome and somatic mutations in single hematopoietic stem and progenitor cells (HSPCs). We found that the fraction of JAK2-mutant HSPCs varied significantly across different myeloid cell types within the same patient. The erythroid progenitor cells were often entirely JAK2-mutant, even when the peripheral blood JAK2-V617F allele burden was low. The novel biological insights uncovered by this work have implications for the prevention and treatment of MPN, as well as the accurate assessment of disease burden in patients. The technology platforms and computational frameworks developed here are broadly applicable to other types of hematological malignancies and cancers.

Funder: U.S. Department of Health & Human Services | NIH | National Cancer Institute (NCI), Funder: Ontario Institute for Cancer Research (Institut Ontarien de Recherche sur le Cancer); doi: https://doi.org/10.13039/100012118, Funder: EMBL Member States EU FP7 Programme projects EurocanPlatform (260791) CAGEKID (241669), Funder: European Union’s Framework Programme For Research and Innovation Horizon 2020 under the Marie Sklodowska-Curie grant agreement no. 703543, Funder: Michael & Susan Dell Foundation; Mary K. Chapman Foundation; CCSG Grant P30 CA016672 (Bioinformatics Shared Resource); ITCR U24 CA199461; GDAN U24 CA210949; GDAN U24 CA210950, Funder: European Commission's H2020 Programme, project SOUND, Grant Agreement no 633974, Funder: Spanish Government (SEV 2015-0493) BSC-Lenovo Master Collaboration Agreement (2015), The Pan-Cancer Analysis of Whole Genomes (PCAWG) project generated a vast amount of whole-genome cancer sequencing resource data. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumor types, we provide a user’s guide to the five publicly available online data exploration and visualization tools introduced in the PCAWG marker paper. These tools are ICGC Data Portal, UCSC Xena, Chromothripsis Explorer, Expression Atlas, and PCAWG-Scout. We detail use cases and analyses for each tool, show how they incorporate outside resources from the larger genomics ecosystem, and demonstrate how the tools can be used together to understand the biology of cancers more deeply. Together, the tools enable researchers to query the complex genomic PCAWG data dynamically and integrate external information, enabling and enhancing interpretation.

The application of convolutional neural networks (ConvNets) to harness high-content screening images or 2D compound representations is gaining increasing attention in drug discovery. However, existing applications often require large data sets for training, or sophisticated pretraining schemes. Here, we show using 33 IC50 data sets from ChEMBL 23 that the in vitro activity of compounds on cancer cell lines and protein targets can be accurately predicted on a continuous scale from their Kekulé structure representations alone by extending existing architectures (AlexNet, DenseNet-201, ResNet152 and VGG-19), which were pretrained on unrelated image data sets. We show that the predictive power of the generated models, which just require standard 2D compound representations as input, is comparable to that of Random Forest (RF) models and fully-connected Deep Neural Networks trained on circular (Morgan) fingerprints. Notably, including additional fully-connected layers further increases the predictive power of the ConvNets by up to 10%. Analysis of the predictions generated by RF models and ConvNets shows that by simply averaging the output of the RF models and ConvNets we obtain significantly lower errors in prediction for multiple data sets, although the effect size is small, than those obtained with either model alone, indicating that the features extracted by the convolutional layers of the ConvNets provide complementary predictive signal to Morgan fingerprints. Lastly, we show that multi-task ConvNets trained on compound images permit to model COX isoform selectivity on a continuous scale with errors in prediction comparable to the uncertainty of the data. Overall, in this work we present a set of ConvNet architectures for the prediction of compound activity from their Kekulé structure representations with state-of-the-art performance, that require no generation of compound descriptors or use of sophisticated image processing techniques. The code needed to reproduce the results presented in this study and all the data sets are provided at https://github.com/isidroc/kekulescope.

Chromothripsis is a mutational phenomenon characterized by massive, clustered genomic rearrangements that occurs in cancer and other diseases. Recent studies in selected cancer types have suggested that chromothripsis may be more common than initially inferred from low-resolution copy-number data. Here, as part of the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA), we analyze patterns of chromothripsis across 2,658 tumors from 38 cancer types using whole-genome sequencing data. We find that chromothripsis events are pervasive across cancers, with a frequency of more than 50% in several cancer types. Whereas canonical chromothripsis profiles display oscillations between two copy-number states, a considerable fraction of events involve multiple chromosomes and additional structural alterations. In addition to non-homologous end joining, we detect signatures of replication-associated processes and templated insertions. Chromothripsis contributes to oncogene amplification and to inactivation of genes such as mismatch-repair-related genes. These findings show that chromothripsis is a major process that drives genome evolution in human cancer.

The Genomics of Malignant Peripheral Nerve Sheath Tumor (GeM) Consortium is an international collaboration focusing on multi-omic analysis of malignant peripheral nerve sheath tumors (MPNSTs), the most aggressive tumor associated with neurofibromatosis type 1 (NF1). Here we present a summary of current knowledge gaps, a description of our consortium and the cohort we have assembled, and an overview of our plans for multi-omic analysis of these tumors. We propose that our analysis will lead to a better understanding of the order and timing of genetic events related to MPNST initiation and progression. Our ten institutions have assembled 96 fresh frozen NF1-related (63%) and sporadic MPNST specimens from 86 subjects with corresponding clinical and pathological data. Clinical data have been collected as part of the International MPNST Registry. We will characterize these tumors with bulk whole genome sequencing, RNAseq, and DNA methylation profiling. In addition, we will perform multiregional analysis and temporal sampling, with the same methodologies, on a subset of nine subjects with NF1-related MPNSTs to assess tumor heterogeneity and cancer evolution. Subsequent multi-omic analyses of additional archival specimens will include deep exome sequencing (500×) and high density copy number arrays for both validation of results based on fresh frozen tumors, and to assess further tumor heterogeneity and evolution. Digital pathology images are being collected in a cloud-based platform for consensus review. The result of these efforts will be the largest MPNST multi-omic dataset with correlated clinical and pathological information ever assembled.

Deficient DNA mismatch repair (dMMR) in human cancer is associated with high tumor mutation burden (TMB), frameshift mutation-derived neoantigens, increased T cell infiltration, and remarkable responsiveness to immune checkpoint blockade (ICB) therapy. Nevertheless, about half of these tumors do not respond to ICB for unclear reasons. While tumor cell line transplant models of dMMR have helped solidify the importance of TMB in immune response, critical questions remain regarding the role of immune surveillance in the evolution of dMMR tumors induced in vivo. Here, we developed autochthonous mouse models of lung and colon cancer with highly efficient ablation of MMR genes via in vivo CRISPR/Cas9 targeting. Surprisingly, dMMR in these models did not result in increased immunogenicity or response to ICB, which we showed is driven by profound intratumoral heterogeneity. Studies in animals depleted of T cells further demonstrated that immune surveillance in dMMR tumors has no impact on TMB but shapes the clonal architecture of neoantigens. These results provide important context for understanding immune evasion in cancers with high TMB and have major implications for therapies aimed at increasing TMB. Citation Format: Peter M. K. Westcott, Francesc M. Remolar, Olivia Smith, Haley Hauck, Nathan J. Sacks, Zackery A. Ely, Alex M. Jaeger, William M. Rideout, Arjun Bhutkar, Daniel Zhang, Mary C. Beytagh, Roderick T. Bronson, David A. Canner, Santiago Naranjo, Abbey Jin, J. J. Patten, Amanda M. Cruz, Isidro Cortes-Ciriano, Tyler Jacks. Mismatch repair deficiency is not sufficient to increase tumor immunogenicity [abstract]. In: Abstracts: AACR Virtual Special Conference: Tumor Immunology and Immunotherapy; 2021 Oct 5-6. Philadelphia (PA): AACR; Cancer Immunol Res 2022;10(1 Suppl):Abstract nr P023.

Making predictions with an associated confidence is highly desirable as it facilitates decision making and resource prioritization. Conformal regression is a machine learning framework that allows the user to define the required confidence and delivers predictions that are guaranteed to be correct to the selected extent. In this study, we apply conformal regression to model molecular properties and bioactivity values and investigate different ways to scale the resultant prediction intervals to create as efficient (i.e., narrow) regressors as possible. Different algorithms to estimate the prediction uncertainty were used to normalize the prediction ranges, and the different approaches were evaluated on 29 publicly available data sets. Our results show that the most efficient conformal regressors are obtained when using the natural exponential of the ensemble standard deviation from the underlying random forest to scale the prediction intervals, but other approaches were almost as efficient. This approach afforded an average prediction range of 1.65 pIC50 units at the 80% confidence level when applied to bioactivity modeling. The choice of nonconformity function has a pronounced impact on the average prediction range with a difference of close to one log unit in bioactivity between the tightest and widest prediction range. Overall, conformal regression is a robust approach to generate bioactivity predictions with associated confidence.