Quality control stress test for deep learning-based diagnostic model in digital pathology.

Digital pathology provides a possibility for computational analysis of histological slides and automatization of routine pathological tasks. Histological slides are very heterogeneous concerning staining, sections' thickness, and artifacts arising during tissue processing, cutting, staining, and digitization. In this study, we digitally reproduce major types of artifacts. Using six datasets from four different institutions digitized by different scanner systems, we systematically explore artifacts' influence on the accuracy of the pre-trained, validated, deep learning-based model for prostate cancer detection in histological slides. We provide evidence that any histological artifact dependent on severity can lead to a substantial loss in model performance. Strategies for the prevention of diagnostic model accuracy losses in the context of artifacts are warranted. Stress-testing of diagnostic models using synthetically generated artifacts might be an essential step during clinical validation of deep learning-based algorithms.

Telomerase activation by genomic rearrangements in high-risk neuroblastoma.

Neuroblastoma is a malignant paediatric tumour of the sympathetic nervous system. Roughly half of these tumours regress spontaneously or are cured by limited therapy. By contrast, high-risk neuroblastomas have an unfavourable clinical course despite intensive multimodal treatment, and their molecular basis has remained largely elusive. Here we have performed whole-genome sequencing of 56 neuroblastomas (high-risk, n = 39; low-risk, n = 17) and discovered recurrent genomic rearrangements affecting a chromosomal region at 5p15.33 proximal of the telomerase reverse transcriptase gene (TERT). These rearrangements occurred only in high-risk neuroblastomas (12/39, 31%) in a mutually exclusive fashion with MYCN amplifications and ATRX mutations, which are known genetic events in this tumour type. In an extended case series (n = 217), TERT rearrangements defined a subgroup of high-risk tumours with particularly poor outcome. Despite a large structural diversity of these rearrangements, they all induced massive transcriptional upregulation of TERT. In the remaining high-risk tumours, TERT expression was also elevated in MYCN-amplified tumours, whereas alternative lengthening of telomeres was present in neuroblastomas without TERT or MYCN alterations, suggesting that telomere lengthening represents a central mechanism defining this subtype. The 5p15.33 rearrangements juxtapose the TERT coding sequence to strong enhancer elements, resulting in massive chromatin remodelling and DNA methylation of the affected region. Supporting a functional role of TERT, neuroblastoma cell lines bearing rearrangements or amplified MYCN exhibited both upregulated TERT expression and enzymatic telomerase activity. In summary, our findings show that remodelling of the genomic context abrogates transcriptional silencing of TERT in high-risk neuroblastoma and places telomerase activation in the centre of transformation in a large fraction of these tumours.

iRODS metadata management for a cancer genome analysis workflow.

BACKGROUND: The massive amounts of data from next generation sequencing (NGS) methods pose various challenges with respect to data security, storage and metadata management. While there is a broad range of data analysis pipelines, these challenges remain largely unaddressed to date. RESULTS: We describe the integration of the open-source metadata management system iRODS (Integrated Rule-Oriented Data System) with a cancer genome analysis pipeline in a high performance computing environment. The system allows for customized metadata attributes as well as fine-grained protection rules and is augmented by a user-friendly front-end for metadata input. This results in a robust, efficient end-to-end workflow under consideration of data security, central storage and unified metadata information. CONCLUSIONS: Integrating iRODS with an NGS data analysis pipeline is a suitable method for addressing the challenges of data security, storage and metadata management in NGS environments.

A phylogenomic approach to resolve the arthropod tree of life.

Arthropods were the first animals to conquer land and air. They encompass more than three quarters of all described living species. This extraordinary evolutionary success is based on an astoundingly wide array of highly adaptive body organizations. A lack of robustly resolved phylogenetic relationships, however, currently impedes the reliable reconstruction of the underlying evolutionary processes. Here, we show that phylogenomic data can substantially advance our understanding of arthropod evolution and resolve several conflicts among existing hypotheses. We assembled a data set of 233 taxa and 775 genes from which an optimally informative data set of 117 taxa and 129 genes was finally selected using new heuristics and compared with the unreduced data set. We included novel expressed sequence tag (EST) data for 11 species and all published phylogenomic data augmented by recently published EST data on taxonomically important arthropod taxa. This thorough sampling reduces the chance of obtaining spurious results due to stochastic effects of undersampling taxa and genes. Orthology prediction of genes, alignment masking tools, and selection of most informative genes due to a balanced taxa-gene ratio using new heuristics were established. Our optimized data set robustly resolves major arthropod relationships. We received strong support for a sister group relationship of onychophorans and euarthropods and strong support for a close association of tardigrades and cycloneuralia. Within pancrustaceans, our analyses yielded paraphyletic crustaceans and monophyletic hexapods and robustly resolved monophyletic endopterygote insects. However, our analyses also showed for few deep splits that were recently thought to be resolved, for example, the position of myriapods, a remarkable sensitivity to methods of analyses.

CaMuS: simultaneous fitting and de novo imputation of cancer mutational signature.

The identification of the mutational processes operating in tumour cells has implications for cancer diagnosis and therapy. These processes leave mutational patterns on the cancer genomes, which are referred to as mutational signatures. Recently, 81 mutational signatures have been inferred using computational algorithms on sequencing data of 23,879 samples. However, these published signatures may not always offer a comprehensive view on the biological processes underlying tumour types that are not included or underrepresented in the reference studies. To circumvent this problem, we designed CaMuS (Cancer Mutational Signatures) to construct de novo signatures while simultaneously fitting publicly available mutational signatures. Furthermore, we propose to estimate signature similarity by comparing probability distributions using the Hellinger distance. We applied CaMuS to infer signatures of mutational processes in poorly studied cancer types. We used whole genome sequencing data of 56 neuroblastoma, thus providing evidence for the versatility of CaMuS. Using simulated data, we compared the performance of CaMuS to sigfit, a recently developed algorithm with comparable inference functionalities. CaMuS and sigfit reconstructed the simulated datasets with similar accuracy; however two main features may argue for CaMuS over sigfit: (i) superior computational performance and (ii) a reliable parameter selection method to avoid spurious signatures.

Copy-number analysis and inference of subclonal populations in cancer genomes using Sclust.

The genomes of cancer cells constantly change during pathogenesis. This evolutionary process can lead to the emergence of drug-resistant mutations in subclonal populations, which can hinder therapeutic intervention in patients. Data derived from massively parallel sequencing can be used to infer these subclonal populations using tumor-specific point mutations. The accurate determination of copy-number changes and tumor impurity is necessary to reliably infer subclonal populations by mutational clustering. This protocol describes how to use Sclust, a copy-number analysis method with a recently developed mutational clustering approach. In a series of simulations and comparisons with alternative methods, we have previously shown that Sclust accurately determines copy-number states and subclonal populations. Performance tests show that the method is computationally efficient, with copy-number analysis and mutational clustering taking <10 min. Sclust is designed such that even non-experts in computational biology or bioinformatics with basic knowledge of the Linux/Unix command-line syntax should be able to carry out analyses of subclonal populations.

Clonal dynamics towards the development of venetoclax resistance in chronic lymphocytic leukemia.

Deciphering the evolution of cancer cells under therapeutic pressure is a crucial step to understand the mechanisms that lead to treatment resistance. To this end, we analyzed whole-exome sequencing data of eight chronic lymphocytic leukemia (CLL) patients that developed resistance upon BCL2-inhibition by venetoclax. Here, we report recurrent mutations in BTG1 (2 patients) and homozygous deletions affecting CDKN2A/B (3 patients) that developed during treatment, as well as a mutation in BRAF and a high-level focal amplification of CD274 (PD-L1) that might pinpoint molecular aberrations offering structures for further therapeutic interventions.

A mechanistic classification of clinical phenotypes in neuroblastoma.

Neuroblastoma is a pediatric tumor of the sympathetic nervous system. Its clinical course ranges from spontaneous tumor regression to fatal progression. To investigate the molecular features of the divergent tumor subtypes, we performed genome sequencing on 416 pretreatment neuroblastomas and assessed telomere maintenance mechanisms in 208 of these tumors. We found that patients whose tumors lacked telomere maintenance mechanisms had an excellent prognosis, whereas the prognosis of patients whose tumors harbored telomere maintenance mechanisms was substantially worse. Survival rates were lowest for neuroblastoma patients whose tumors harbored telomere maintenance mechanisms in combination with RAS and/or p53 pathway mutations. Spontaneous tumor regression occurred both in the presence and absence of these mutations in patients with telomere maintenance-negative tumors. On the basis of these data, we propose a mechanistic classification of neuroblastoma that may benefit the clinical management of patients.

Integrative genomic profiling of large-cell neuroendocrine carcinomas reveals distinct subtypes of high-grade neuroendocrine lung tumors.

Pulmonary large-cell neuroendocrine carcinomas (LCNECs) have similarities with other lung cancers, but their precise relationship has remained unclear. Here we perform a comprehensive genomic (n = 60) and transcriptomic (n = 69) analysis of 75 LCNECs and identify two molecular subgroups: "type I LCNECs" with bi-allelic TP53 and STK11/KEAP1 alterations (37%), and "type II LCNECs" enriched for bi-allelic inactivation of TP53 and RB1 (42%). Despite sharing genomic alterations with adenocarcinomas and squamous cell carcinomas, no transcriptional relationship was found; instead LCNECs form distinct transcriptional subgroups with closest similarity to SCLC. While type I LCNECs and SCLCs exhibit a neuroendocrine profile with ASCL1high/DLL3high/NOTCHlow, type II LCNECs bear TP53 and RB1 alterations and differ from most SCLC tumors with reduced neuroendocrine markers, a pattern of ASCL1low/DLL3low/NOTCHhigh, and an upregulation of immune-related pathways. In conclusion, LCNECs comprise two molecularly defined subgroups, and distinguishing them from SCLC may allow stratified targeted treatment of high-grade neuroendocrine lung tumors.

Leveraging the power of high performance computing for next generation sequencing data analysis: tricks and twists from a high throughput exome workflow.

Next generation sequencing (NGS) has been a great success and is now a standard method of research in the life sciences. With this technology, dozens of whole genomes or hundreds of exomes can be sequenced in rather short time, producing huge amounts of data. Complex bioinformatics analyses are required to turn these data into scientific findings. In order to run these analyses fast, automated workflows implemented on high performance computers are state of the art. While providing sufficient compute power and storage to meet the NGS data challenge, high performance computing (HPC) systems require special care when utilized for high throughput processing. This is especially true if the HPC system is shared by different users. Here, stability, robustness and maintainability are as important for automated workflows as speed and throughput. To achieve all of these aims, dedicated solutions have to be developed. In this paper, we present the tricks and twists that we utilized in the implementation of our exome data processing workflow. It may serve as a guideline for other high throughput data analysis projects using a similar infrastructure. The code implementing our solutions is provided in the supporting information files.

Frequent mutations in chromatin-remodelling genes in pulmonary carcinoids.

Pulmonary carcinoids are rare neuroendocrine tumours of the lung. The molecular alterations underlying the pathogenesis of these tumours have not been systematically studied so far. Here we perform gene copy number analysis (n=54), genome/exome (n=44) and transcriptome (n=69) sequencing of pulmonary carcinoids and observe frequent mutations in chromatin-remodelling genes. Covalent histone modifiers and subunits of the SWI/SNF complex are mutated in 40 and 22.2% of the cases, respectively, with MEN1, PSIP1 and ARID1A being recurrently affected. In contrast to small-cell lung cancer and large-cell neuroendocrine lung tumours, TP53 and RB1 mutations are rare events, suggesting that pulmonary carcinoids are not early progenitor lesions of the highly aggressive lung neuroendocrine tumours but arise through independent cellular mechanisms. These data also suggest that inactivation of chromatin-remodelling genes is sufficient to drive transformation in pulmonary carcinoids.