Deep learning is a subdiscipline of artificial intelligence that uses a machine learning technique called artificial neural networks to extract patterns and make predictions from large data sets. The ...increasing adoption of deep learning across healthcare domains together with the availability of highly characterised cancer datasets has accelerated research into the utility of deep learning in the analysis of the complex biology of cancer. While early results are promising, this is a rapidly evolving field with new knowledge emerging in both cancer biology and deep learning. In this review, we provide an overview of emerging deep learning techniques and how they are being applied to oncology. We focus on the deep learning applications for omics data types, including genomic, methylation and transcriptomic data, as well as histopathology-based genomic inference, and provide perspectives on how the different data types can be integrated to develop decision support tools. We provide specific examples of how deep learning may be applied in cancer diagnosis, prognosis and treatment management. We also assess the current limitations and challenges for the application of deep learning in precision oncology, including the lack of phenotypically rich data and the need for more explainable deep learning models. Finally, we conclude with a discussion of how current obstacles can be overcome to enable future clinical utilisation of deep learning.
For complex machine learning (ML) algorithms to gain widespread acceptance in decision making, we must be able to identify the features driving the predictions. Explainability models allow ...transparency of ML algorithms, however their reliability within high-dimensional data is unclear. To test the reliability of the explainability model SHapley Additive exPlanations (SHAP), we developed a convolutional neural network to predict tissue classification from Genotype-Tissue Expression (GTEx) RNA-seq data representing 16,651 samples from 47 tissues. Our classifier achieved an average F1 score of 96.1% on held-out GTEx samples. Using SHAP values, we identified the 2423 most discriminatory genes, of which 98.6% were also identified by differential expression analysis across all tissues. The SHAP genes reflected expected biological processes involved in tissue differentiation and function. Moreover, SHAP genes clustered tissue types with superior performance when compared to all genes, genes detected by differential expression analysis, or random genes. We demonstrate the utility and reliability of SHAP to explain a deep learning model and highlight the strengths of applying ML to transcriptome data.
Acquired resistance to PARP inhibitors (PARPi) is a major challenge for the clinical management of high grade serous ovarian cancer (HGSOC). Here, we demonstrate CX-5461, the first-in-class inhibitor ...of RNA polymerase I transcription of ribosomal RNA genes (rDNA), induces replication stress and activates the DNA damage response. CX-5461 co-operates with PARPi in exacerbating replication stress and enhances therapeutic efficacy against homologous recombination (HR) DNA repair-deficient HGSOC-patient-derived xenograft (PDX) in vivo. We demonstrate CX-5461 has a different sensitivity spectrum to PARPi involving MRE11-dependent degradation of replication forks. Importantly, CX-5461 exhibits in vivo single agent efficacy in a HGSOC-PDX with reduced sensitivity to PARPi by overcoming replication fork protection. Further, we identify CX-5461-sensitivity gene expression signatures in primary and relapsed HGSOC. We propose CX-5461 is a promising therapy in combination with PARPi in HR-deficient HGSOC and also as a single agent for the treatment of relapsed disease.
Abstract
Cells within the tumour microenvironment (TME) can impact tumour development and influence treatment response. Computational approaches have been developed to deconvolve the TME from bulk ...RNA-seq. Using scRNA-seq profiling from breast tumours we simulate thousands of bulk mixtures, representing tumour purities and cell lineages, to compare the performance of nine TME deconvolution methods (BayesPrism, Scaden, CIBERSORTx, MuSiC, DWLS, hspe, CPM, Bisque, and EPIC).
S
ome methods are more robust in deconvolving mixtures with high tumour purity levels. Most methods tend to mis-predict normal epithelial for cancer epithelial as tumour purity increases, a finding that is validated in two independent datasets. The breast cancer molecular subtype influences this mis-prediction. BayesPrism and DWLS have the lowest combined numbers of false positives and false negatives, and have the best performance when deconvolving granular immune lineages. Our findings highlight the need for more single-cell characterisation of rarer cell types, and suggest that tumour cell compositions should be considered when deconvolving the TME.
Uncertainty estimation is crucial for understanding the reliability of deep learning (DL) predictions, and critical for deploying DL in the clinic. Differences between training and production ...datasets can lead to incorrect predictions with underestimated uncertainty. To investigate this pitfall, we benchmarked one pointwise and three approximate Bayesian DL models for predicting cancer of unknown primary, using three RNA-seq datasets with 10,968 samples across 57 cancer types. Our results highlight that simple and scalable Bayesian DL significantly improves the generalisation of uncertainty estimation. Moreover, we designed a prototypical metric-the area between development and production curve (ADP), which evaluates the accuracy loss when deploying models from development to production. Using ADP, we demonstrate that Bayesian DL improves accuracy under data distributional shifts when utilising 'uncertainty thresholding'. In summary, Bayesian DL is a promising approach for generalising uncertainty, improving performance, transparency, and safety of DL models for deployment in the real world.
Genetic and epigenetic factors contribute to the development of cancer. Epigenetic dysregulation is common in gynaecological cancers and includes altered methylation at CpG islands in gene promoter ...regions, global demethylation that leads to genome instability and histone modifications. Histones are a major determinant of chromosomal conformation and stability, and unlike DNA methylation, which is generally associated with gene silencing, are amenable to post-translational modifications that induce facultative chromatin regions, or condensed transcriptionally silent regions that decondense resulting in global alteration of gene expression. In comparison, other components, crucial to the manipulation of chromatin dynamics, such as histone modifying enzymes, are not as well-studied. Inhibitors targeting DNA modifying enzymes, particularly histone modifying enzymes represent a potential cancer treatment. Due to the ability of epigenetic therapies to target multiple pathways simultaneously, tumours with complex mutational landscapes affected by multiple driver mutations may be most amenable to this type of inhibitor. Interrogation of the actionable landscape of different gynaecological cancer types has revealed that some patients have biomarkers which indicate potential sensitivity to epigenetic inhibitors. In this review we describe the role of epigenetics in gynaecological cancers and highlight how it may exploited for treatment.
Epigenetic alterations, including aberrant DNA methylation, are now recognized as
hallmarks of cancer, which can contribute to cancer initiation, progression, therapy responses and therapy ...resistance. Methylation of gene promoters can have a range of impacts on cancer risk, clinical stratification and therapeutic outcomes. We provide several important examples of genes, which can be silenced or activated by promoter methylation and highlight their clinical implications. These include the mismatch DNA repair genes
and
, homologous recombination DNA repair genes
and
, the
oncogene and genes within the
tumour suppressor axis. We also discuss how these methylation changes might occur in the first place - whether in the context of the CpG island methylator phenotype or constitutional DNA methylation. The choice of assay used to measure methylation can have a significant impact on interpretation of methylation states, and some examples where this can influence clinical decision-making are presented. Aberrant DNA methylation patterns in circulating tumour DNA (ctDNA) are also showing great promise in the context of non-invasive cancer detection and monitoring using liquid biopsies; however, caution must be taken in interpreting these results in cases where constitutional methylation may be present. Thus, this review aims to provide researchers and clinicians with a comprehensive summary of this broad, but important subject, illustrating the potentials and pitfalls of assessing aberrant DNA methylation in cancer.
Endometrial cancer (EC) is a major gynecological cancer with increasing incidence. It comprises four molecular subtypes with differing etiology, prognoses, and responses to chemotherapy. In the ...future, clinical trials testing new single agents or combination therapies will be targeted to the molecular subtype most likely to respond. As pre-clinical models that faithfully represent the molecular subtypes of EC are urgently needed, we sought to develop and characterize a panel of novel EC patient-derived xenograft (PDX) models.
Here, we report whole exome or whole genome sequencing of 11 PDX models and their matched primary tumor. Analysis of multiple PDX lineages and passages was performed to study tumor heterogeneity across lineages and/or passages. Based on recent reports of frequent defects in the homologous recombination (HR) pathway in EC, we assessed mutational signatures and HR deficiency scores and correlated these with in vivo responses to the PARP inhibitor (PARPi) talazoparib in six PDXs representing the copy number high/p53-mutant and mismatch-repair deficient molecular subtypes of EC.
PDX models were successfully generated from grade 2/3 tumors, including three uterine carcinosarcomas. The models showed similar histomorphology to the primary tumors and represented all four molecular subtypes of EC, including five mismatch-repair deficient models. The different PDX lineages showed a wide range of inter-tumor and intra-tumor heterogeneity. However, for most PDX models, one arm recapitulated the molecular landscape of the primary tumor without major genomic drift. An in vivo response to talazoparib was detected in four copy number high models. Two models (carcinosarcomas) showed a response consistent with stable disease and two models (one copy number high serous EC and another carcinosarcoma) showed significant tumor growth inhibition, albeit one consistent with progressive disease; however, all lacked the HR deficiency genomic signature.
EC PDX models represent the four molecular subtypes of disease and can capture intra-tumor heterogeneity of the original primary tumor. PDXs of the copy number high molecular subtype showed sensitivity to PARPi; however, deeper and more durable responses will likely require combination of PARPi with other agents.
Risk of endometrial cancer (EC) is increased ~2-fold for women with a family history of cancer, partly due to inherited pathogenic variants in mismatch repair (MMR) genes. We explored the role of ...additional genes as explanation for familial EC presentation by investigating germline and EC tumor sequence data from The Cancer Genome Atlas (
= 539; 308 European ancestry), and germline data from 33 suspected familial European ancestry EC patients demonstrating immunohistochemistry-detected tumor MMR proficiency. Germline variants in MMR and 26 other known/candidate EC risk genes were annotated for pathogenicity in the two EC datasets, and also for European ancestry individuals from gnomAD as a population reference set (
= 59,095). Ancestry-matched case-control comparisons of germline variant frequency and/or sequence data from suspected familial EC cases highlighted
,
,
,
and
as candidates for large-scale risk association studies. Tumor mutational signature analysis identified a microsatellite-high signature for all cases with a germline pathogenic MMR gene variant. Signature analysis also indicated that germline loss-of-function variants in homologous recombination (
,
,
) or base excision (
,
) repair genes can contribute to EC development in some individuals with germline variants in these genes. These findings have implications for expanded therapeutic options for EC cases.
Background:
Despite initial response to platinum-based chemotherapy and PARP inhibitor therapy (PARPi), nearly all recurrent high-grade serous ovarian cancer (HGSC) will acquire lethal drug ...resistance; indeed, ~15% of individuals have de novo platinum-refractory disease.
Objectives:
To determine the potential of anti-microtubule agent (AMA) therapy (paclitaxel, vinorelbine and eribulin) in platinum-resistant or refractory (PRR) HGSC by assessing response in patient-derived xenograft (PDX) models of HGSC.
Design and methods:
Of 13 PRR HGSC PDX, six were primary PRR, derived from chemotherapy-naïve samples (one was BRCA2 mutant) and seven were from samples obtained following chemotherapy treatment in the clinic (five were mutant for either BRCA1 or BRCA2 (BRCA1/2), four with prior PARPi exposure), recapitulating the population of individuals with aggressive treatment-resistant HGSC in the clinic. Molecular analyses and in vivo treatment studies were undertaken.
Results:
Seven out of thirteen PRR PDX (54%) were sensitive to treatment with the AMA, eribulin (time to progressive disease (PD) ⩾100 days from the start of treatment) and 11 out of 13 PDX (85%) derived significant benefit from eribulin time to harvest (TTH) for each PDX with p < 0.002. In 5 out of 10 platinum-refractory HGSC PDX (50%) and one out of three platinum-resistant PDX (33%), eribulin was more efficacious than was cisplatin, with longer time to PD and significantly extended TTH (each PDX p < 0.02). Furthermore, four of these models were extremely sensitive to all three AMA tested, maintaining response until the end of the experiment (120d post-treatment start). Despite harbouring secondary BRCA2 mutations, two BRCA2-mutant PDX models derived from heavily pre-treated individuals were sensitive to AMA. PRR HGSC PDX models showing greater sensitivity to AMA had high proliferative indices and oncogene expression. Two PDX models, both with prior chemotherapy and/or PARPi exposure, were refractory to all AMA, one of which harboured the SLC25A40-ABCB1 fusion, known to upregulate drug efflux via MDR1.
Conclusion:
The efficacy observed for eribulin in PRR HGSC PDX was similar to that observed for paclitaxel, which transformed ovarian cancer clinical practice. Eribulin is therefore worthy of further consideration in clinical trials, particularly in ovarian carcinoma with early failure of carboplatin/paclitaxel chemotherapy.