Innate acting memory Th1 cells modulate heterologous diseases.

Through immune memory, infections have a lasting effect on the host. While memory cells enable accelerated and enhanced responses upon rechallenge with the same pathogen, their impact on susceptibility to unrelated diseases is unclear. We identify a subset of memory T helper 1 (Th1) cells termed innate acting memory T (TIA) cells that originate from a viral infection and produce IFN-γ with innate kinetics upon heterologous challenge in vivo. Activation of memory TIA cells is induced in response to IL-12 in combination with IL-18 or IL-33 but is TCR independent. Rapid IFN-γ production by memory TIA cells is protective in subsequent heterologous challenge with the bacterial pathogen Legionella pneumophila. In contrast, antigen-independent reactivation of CD4+ memory TIA cells accelerates disease onset in an autoimmune model of multiple sclerosis. Our findings demonstrate that memory Th1 cells can acquire additional TCR-independent functionality to mount rapid, innate-like responses that modulate susceptibility to heterologous challenges.

In vivo identification of tumor- suppressive PTEN ceRNAs in an oncogenic BRAF-induced mouse model of melanoma.

We recently proposed that competitive endogenous RNAs (ceRNAs) sequester microRNAs to regulate mRNA transcripts containing common microRNA recognition elements (MREs). However, the functional role of ceRNAs in cancer remains unknown. Loss of PTEN, a tumor suppressor regulated by ceRNA activity, frequently occurs in melanoma. Here, we report the discovery of significant enrichment of putative PTEN ceRNAs among genes whose loss accelerates tumorigenesis following Sleeping Beauty insertional mutagenesis in a mouse model of melanoma. We validated several putative PTEN ceRNAs and further characterized one, the ZEB2 transcript. We show that ZEB2 modulates PTEN protein levels in a microRNA-dependent, protein coding-independent manner. Attenuation of ZEB2 expression activates the PI3K/AKT pathway, enhances cell transformation, and commonly occurs in human melanomas and other cancers expressing low PTEN levels. Our study genetically identifies multiple putative microRNA decoys for PTEN, validates ZEB2 mRNA as a bona fide PTEN ceRNA, and demonstrates that abrogated ZEB2 expression cooperates with BRAF(V600E) to promote melanomagenesis.

Fragmentstein-facilitating data reuse for cell-free DNA fragment analysis.

SUMMARY: Method development for the analysis of cell-free DNA (cfDNA) sequencing data is impeded by limited data sharing due to the strict control of sensitive genomic data. An existing solution for facilitating data sharing removes nucleotide-level information from raw cfDNA sequencing data, keeping alignment coordinates only. This simplified format can be publicly shared and would, theoretically, suffice for common functional analyses of cfDNA data. However, current bioinformatics software requires nucleotide-level information and cannot process the simplified format. We present Fragmentstein, a command-line tool for converting non-sensitive cfDNA-fragmentation data into alignment mapping (BAM) files. Fragmentstein complements fragment coordinates with sequence information from a reference genome to reconstruct BAM files. We demonstrate the utility of Fragmentstein by showing the feasibility of copy number variant (CNV), nucleosome occupancy, and fragment length analyses from non-sensitive fragmentation data. AVAILABILITY AND IMPLEMENTATION: Implemented in bash, Fragmentstein is available at https://github.com/uzh-dqbm-cmi/fragmentstein, licensed under GNU GPLv3.

COL10A1 expression distinguishes a subset of cancer-associated fibroblasts present in the stroma of high-risk basal cell carcinoma.

BACKGROUND: Basal cell carcinoma (BCC) is the most frequently diagnosed skin cancer and the most common malignancy in humans. Different morphological subtypes of BCC are associated with low- or high-risk of recurrence and aggressiveness, but the underlying biology of how the individual subtypes arise remains largely unknown. Because the majority of BCCs appear to arise from mutations in the same pathway, we hypothesized that BCC development, growth and invasive potential is also influenced by the tumor microenvironment and in particular by cancer-associated fibroblasts (CAFs) and their secreted factors. OBJECTIVE: We aimed to characterize the stroma of the different BCC subtypes with a focus on CAF populations. METHODS: To investigate the stromal features of the different BCC subtypes, we applied laser-capture microdissection (LCM) followed by RNA sequencing. A cohort of 15 BCC samples from 5 different "pure" subtypes (superficial, nodular, micronodular, sclerosing and basosquamous; n=3 each) were selected and included in the analysis. Healthy skin was used as a control (n=6). We confirmed the results by immunohistochemistry. We validated our findings in two independent, public single-cell RNA sequencing (scRNAseq) datasets and by RNAscope. RESULTS: The stroma of the different BCC subtypes have distinct gene expression signatures. Nodular and micronodular seem to have the most similar signatures, while superficial and sclerosing the most different. By comparing low- and high-risk BCC subtypes, we observed that Collagen 10A1 (COL10A1) is overexpressed in the stroma of sclerosing/infiltrative and basosquamous but not micronodular high-risk subtypes. Those findings were confirmed by immunohistochemistry in a cohort of 89 different BCC and 13 healthy skin samples. Moreover, scRNAseq analysis of BCCs of two independent datasets showed that the COL10A1-expressing population of cells is associated with the stroma adjacent to invasive BCC and shows extracellular matrix remodeling features. CONCLUSION: We identified COL10A1 as a marker of high-risk BCC, in particular of the sclerosing/infiltrative and basosquamous subtypes. We demonstrated at the single cell level that COL10A1 is expressed by a specific CAF population associated with the stroma of invasive BCC. This opens up new tailored treatment options as well as a new prognostic biomarker for BCC progression.

Vision transformer assisting rheumatologists in screening for capillaroscopy changes in systemic sclerosis: an artificial intelligence model.

OBJECTIVES: The first objective of this study was to implement and assess the performance and reliability of a vision transformer (ViT)-based deep-learning model, an 'off-the-shelf' artificial intelligence solution, for identifying distinct signs of microangiopathy in nailfold capilloroscopy (NFC) images of patients with SSc. The second objective was to compare the ViT's analysis performance with that of practising rheumatologists. METHODS: NFC images of patients prospectively enrolled in our European Scleroderma Trials and Research group (EUSTAR) and Very Early Diagnosis of Systemic Sclerosis (VEDOSS) local registries were used. The primary outcome investigated was the ViT's classification performance for identifying disease-associated changes (enlarged capillaries, giant capillaries, capillary loss, microhaemorrhages) and the presence of the scleroderma pattern in these images using a cross-fold validation setting. The secondary outcome involved a comparison of the ViT's performance vs that of rheumatologists on a reliability set, consisting of a subset of 464 NFC images with majority vote-derived ground-truth labels. RESULTS: We analysed 17 126 NFC images derived from 234 EUSTAR and 55 VEDOSS patients. The ViT had good performance in identifying the various microangiopathic changes in capillaries by NFC [area under the curve (AUC) from 81.8% to 84.5%]. In the reliability set, the rheumatologists reached a higher average accuracy, as well as a better trade-off between sensitivity and specificity compared with the ViT. However, the annotators' performance was variable, and one out of four rheumatologists showed equal or lower classification measures compared with the ViT. CONCLUSIONS: The ViT is a modern, well-performing and readily available tool for assessing patterns of microangiopathy on NFC images, and it may assist rheumatologists in generating consistent and high-quality NFC reports; however, the final diagnosis of a scleroderma pattern in any individual case needs the judgement of an experienced observer.

AI support for ethical decision-making around resuscitation: proceed with care.

Artificial intelligence (AI) systems are increasingly being used in healthcare, thanks to the high level of performance that these systems have proven to deliver. So far, clinical applications have focused on diagnosis and on prediction of outcomes. It is less clear in what way AI can or should support complex clinical decisions that crucially depend on patient preferences. In this paper, we focus on the ethical questions arising from the design, development and deployment of AI systems to support decision-making around cardiopulmonary resuscitation and the determination of a patient's Do Not Attempt to Resuscitate status (also known as code status). The COVID-19 pandemic has made us keenly aware of the difficulties physicians encounter when they have to act quickly in stressful situations without knowing what their patient would have wanted. We discuss the results of an interview study conducted with healthcare professionals in a university hospital aimed at understanding the status quo of resuscitation decision processes while exploring a potential role for AI systems in decision-making around code status. Our data suggest that (1) current practices are fraught with challenges such as insufficient knowledge regarding patient preferences, time pressure and personal bias guiding care considerations and (2) there is considerable openness among clinicians to consider the use of AI-based decision support. We suggest a model for how AI can contribute to improve decision-making around resuscitation and propose a set of ethically relevant preconditions-conceptual, methodological and procedural-that need to be considered in further development and implementation efforts.

AttentionDDI: Siamese attention-based deep learning method for drug-drug interaction predictions.

BACKGROUND: Drug-drug interactions (DDIs) refer to processes triggered by the administration of two or more drugs leading to side effects beyond those observed when drugs are administered by themselves. Due to the massive number of possible drug pairs, it is nearly impossible to experimentally test all combinations and discover previously unobserved side effects. Therefore, machine learning based methods are being used to address this issue. METHODS: We propose a Siamese self-attention multi-modal neural network for DDI prediction that integrates multiple drug similarity measures that have been derived from a comparison of drug characteristics including drug targets, pathways and gene expression profiles. RESULTS: Our proposed DDI prediction model provides multiple advantages: (1) It is trained end-to-end, overcoming limitations of models composed of multiple separate steps, (2) it offers model explainability via an Attention mechanism for identifying salient input features and (3) it achieves similar or better prediction performance (AUPR scores ranging from 0.77 to 0.92) compared to state-of-the-art DDI models when tested on various benchmark datasets. Novel DDI predictions are further validated using independent data resources. CONCLUSIONS: We find that a Siamese multi-modal neural network is able to accurately predict DDIs and that an Attention mechanism, typically used in the Natural Language Processing domain, can be beneficially applied to aid in DDI model explainability.

Analyzing Patient Trajectories With Artificial Intelligence.

In digital medicine, patient data typically record health events over time (eg, through electronic health records, wearables, or other sensing technologies) and thus form unique patient trajectories. Patient trajectories are highly predictive of the future course of diseases and therefore facilitate effective care. However, digital medicine often uses only limited patient data, consisting of health events from only a single or small number of time points while ignoring additional information encoded in patient trajectories. To analyze such rich longitudinal data, new artificial intelligence (AI) solutions are needed. In this paper, we provide an overview of the recent efforts to develop trajectory-aware AI solutions and provide suggestions for future directions. Specifically, we examine the implications for developing disease models from patient trajectories along the typical workflow in AI: problem definition, data processing, modeling, evaluation, and interpretation. We conclude with a discussion of how such AI solutions will allow the field to build robust models for personalized risk scoring, subtyping, and disease pathway discovery.

AutoDiscern: rating the quality of online health information with hierarchical encoder attention-based neural networks.

BACKGROUND: Patients increasingly turn to search engines and online content before, or in place of, talking with a health professional. Low quality health information, which is common on the internet, presents risks to the patient in the form of misinformation and a possibly poorer relationship with their physician. To address this, the DISCERN criteria (developed at University of Oxford) are used to evaluate the quality of online health information. However, patients are unlikely to take the time to apply these criteria to the health websites they visit. METHODS: We built an automated implementation of the DISCERN instrument (Brief version) using machine learning models. We compared the performance of a traditional model (Random Forest) with that of a hierarchical encoder attention-based neural network (HEA) model using two language embeddings, BERT and BioBERT. RESULTS: The HEA BERT and BioBERT models achieved average F1-macro scores across all criteria of 0.75 and 0.74, respectively, outperforming the Random Forest model (average F1-macro = 0.69). Overall, the neural network based models achieved 81% and 86% average accuracy at 100% and 80% coverage, respectively, compared to 94% manual rating accuracy. The attention mechanism implemented in the HEA architectures not only provided 'model explainability' by identifying reasonable supporting sentences for the documents fulfilling the Brief DISCERN criteria, but also boosted F1 performance by 0.05 compared to the same architecture without an attention mechanism. CONCLUSIONS: Our research suggests that it is feasible to automate online health information quality assessment, which is an important step towards empowering patients to become informed partners in the healthcare process.

Prototyping a precision oncology 3.0 rapid learning platform.

BACKGROUND: We describe a prototype implementation of a platform that could underlie a Precision Oncology Rapid Learning system. RESULTS: We describe the prototype platform, and examine some important issues and details. In the Appendix we provide a complete walk-through of the prototype platform. CONCLUSIONS: The design choices made in this implementation rest upon ten constitutive hypotheses, which, taken together, define a particular view of how a rapid learning medical platform might be defined, organized, and implemented.

PySeqLab: an open source Python package for sequence labeling and segmentation.

MOTIVATION: Text and genomic data are composed of sequential tokens, such as words and nucleotides that give rise to higher order syntactic constructs. In this work, we aim at providing a comprehensive Python library implementing conditional random fields (CRFs), a class of probabilistic graphical models, for robust prediction of these constructs from sequential data. RESULTS: Python Sequence Labeling (PySeqLab) is an open source package for performing supervised learning in structured prediction tasks. It implements CRFs models, that is discriminative models from (i) first-order to higher-order linear-chain CRFs, and from (ii) first-order to higher-order semi-Markov CRFs (semi-CRFs). Moreover, it provides multiple learning algorithms for estimating model parameters such as (i) stochastic gradient descent (SGD) and its multiple variations, (ii) structured perceptron with multiple averaging schemes supporting exact and inexact search using 'violation-fixing' framework, (iii) search-based probabilistic online learning algorithm (SAPO) and (iv) an interface for Broyden-Fletcher-Goldfarb-Shanno (BFGS) and the limited-memory BFGS algorithms. Viterbi and Viterbi A* are used for inference and decoding of sequences. Using PySeqLab, we built models (classifiers) and evaluated their performance in three different domains: (i) biomedical Natural language processing (NLP), (ii) predictive DNA sequence analysis and (iii) Human activity recognition (HAR). State-of-the-art performance comparable to machine-learning based systems was achieved in the three domains without feature engineering or the use of knowledge sources. AVAILABILITY AND IMPLEMENTATION: PySeqLab is available through https://bitbucket.org/A_2/pyseqlab with tutorials and documentation. CONTACT: ahmed.allam@yale.edu or michael.krauthammer@yale.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Spitz nevi and Spitzoid melanomas: exome sequencing and comparison with conventional melanocytic nevi and melanomas.

We performed exome sequencing of 77 melanocytic specimens composed of Spitz nevi (n=29), Spitzoid melanomas (n=27), and benign melanocytic nevi (n=21), and compared the results with published melanoma sequencing data. Our study highlights the prominent similarity between Spitzoid and conventional melanomas with similar copy number changes and high and equal numbers of ultraviolet-induced coding mutations affecting similar driver genes. Mutations in MEN1, PRKAR1A, and DNMT3A in Spitzoid melanomas may indicate involvement of the protein kinase A pathway, or a role of DNA methylation in the disease. Other than activating HRAS variants, there were few additional mutations in Spitz nevi, and few copy number changes other than 11p amplification and chromosome 9 deletions. Similarly, there were no large-scale copy number alterations and few somatic alterations other than activating BRAF or NRAS mutations in conventional nevi. A presumed melanoma driver mutation (IDH1Arg132Cys) was revealed in one of the benign nevi. In conclusion, our exome data show significantly lower somatic mutation burden in both Spitz and conventional nevi compared with their malignant counterparts, and high genetic similarity between Spitzoid and conventional melanoma.

Global copy number profiling of cancer genomes.

UNLABELLED: In this article, we introduce a robust and efficient strategy for deriving global and allele-specific copy number alternations (CNA) from cancer whole exome sequencing data based on Log R ratios and B-allele frequencies. Applying the approach to the analysis of over 200 skin cancer samples, we demonstrate its utility for discovering distinct CNA events and for deriving ancillary information such as tumor purity. AVAILABILITY AND IMPLEMENTATION: https://github.com/xfwang/CLOSE CONTACT: xuefeng.wang@stonybrook.edu or michael.krauthammer@yale.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Integrated analysis of multidimensional omics data on cutaneous melanoma prognosis.

Multiple types of genetic, epigenetic, and genomic changes have been implicated in cutaneous melanoma prognosis. Many of the existing studies are limited in analyzing a single type of omics measurement and cannot comprehensively describe the biological processes underlying prognosis. As a result, the obtained prognostic models may be less satisfactory, and the identified prognostic markers may be less informative. The recently collected TCGA (The Cancer Genome Atlas) data have a high quality and comprehensive omics measurements, making it possible to more comprehensively and more accurately model prognosis. In this study, we first describe the statistical approaches that can integrate multiple types of omics measurements with the assistance of variable selection and dimension reduction techniques. Data analysis suggests that, for cutaneous melanoma, integrating multiple types of measurements leads to prognostic models with an improved prediction performance. Informative individual markers and pathways are identified, which can provide valuable insights into melanoma prognosis.

Mutadelic: mutation analysis using description logic inferencing capabilities.

MOTIVATION: As next generation sequencing gains a foothold in clinical genetics, there is a need for annotation tools to characterize increasing amounts of patient variant data for identifying clinically relevant mutations. While existing informatics tools provide efficient bulk variant annotations, they often generate excess information that may limit their scalability. RESULTS: We propose an alternative solution based on description logic inferencing to generate workflows that produce only those annotations that will contribute to the interpretation of each variant. Workflows are dynamically generated using a novel abductive reasoning framework called a basic framework for abductive workflow generation (AbFab). Criteria for identifying disease-causing variants in Mendelian blood disorders were identified and implemented as AbFab services. A web application was built allowing users to run workflows generated from the criteria to analyze genomic variants. Significant variants are flagged and explanations provided for why they match or fail to match the criteria. AVAILABILITY AND IMPLEMENTATION: The Mutadelic web application is available for use at http://krauthammerlab.med.yale.edu/mutadelic. CONTACT: michael.krauthammer@yale.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Future perspectives in melanoma research : Meeting report from the "Melanoma Bridge". Napoli, December 1st-4th 2015.

The sixth "Melanoma Bridge Meeting" took place in Naples, Italy, December 1st-4th, 2015. The four sessions at this meeting were focused on: (1) molecular and immune advances; (2) combination therapies; (3) news in immunotherapy; and 4) tumor microenvironment and biomarkers. Recent advances in tumor biology and immunology has led to the development of new targeted and immunotherapeutic agents that prolong progression-free survival (PFS) and overall survival (OS) of cancer patients. Immunotherapies in particular have emerged as highly successful approaches to treat patients with cancer including melanoma, non-small cell lung cancer (NSCLC), renal cell carcinoma (RCC), bladder cancer, and Hodgkin's disease. Specifically, many clinical successes have been using checkpoint receptor blockade, including T cell inhibitory receptors such as cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) and the programmed cell death-1 (PD-1) and its ligand PD-L1. Despite demonstrated successes, responses to immunotherapy interventions occur only in a minority of patients. Attempts are being made to improve responses to immunotherapy by developing biomarkers. Optimizing biomarkers for immunotherapy could help properly select patients for treatment and help to monitor response, progression and resistance that are critical challenges for the immuno-oncology (IO) field. Importantly, biomarkers could help to design rational combination therapies. In addition, biomarkers may help to define mechanism of action of different agents, dose selection and to sequence drug combinations. However, biomarkers and assays development to guide cancer immunotherapy is highly challenging for several reasons: (i) multiplicity of immunotherapy agents with different mechanisms of action including immunotherapies that target activating and inhibitory T cell receptors (e.g., CTLA-4, PD-1, etc.); adoptive T cell therapies that include tissue infiltrating lymphocytes (TILs), chimeric antigen receptors (CARs), and T cell receptor (TCR) modified T cells; (ii) tumor heterogeneity including changes in antigenic profiles over time and location in individual patient; and (iii) a variety of immune-suppressive mechanisms in the tumor microenvironment (TME) including T regulatory cells (Treg), myeloid derived suppressor cells (MDSC) and immunosuppressive cytokines. In addition, complex interaction of tumor-immune system further increases the level of difficulties in the process of biomarkers development and their validation for clinical use. Recent clinical trial results have highlighted the potential for combination therapies that include immunomodulating agents such as anti-PD-1 and anti-CTLA-4. Agents targeting other immune inhibitory (e.g., Tim-3) or immune stimulating (e.g., CD137) receptors on T cells and other approaches such as adoptive cell transfer are tested for clinical efficacy in melanoma as well. These agents are also being tested in combination with targeted therapies to improve upon shorter-term responses thus far seen with targeted therapy. Various locoregional interventions that demonstrate promising results in treatment of advanced melanoma are also integrated with immunotherapy agents and the combinations with cytotoxic chemotherapy and inhibitors of angiogenesis are changing the evolving landscape of therapeutic options and are being evaluated to prevent or delay resistance and to further improve survival rates for melanoma patients' population. This meeting's specific focus was on advances in immunotherapy and combination therapy for melanoma. The importance of understanding of melanoma genomic background for development of novel therapies and biomarkers for clinical application to predict the treatment response was an integral part of the meeting. The overall emphasis on biomarkers supports novel concepts toward integrating biomarkers into personalized-medicine approach for treatment of patients with melanoma across the entire spectrum of disease stage. Translation of the knowledge gained from the biology of tumor microenvironment across different tumors represents a bridge to impact on prognosis and response to therapy in melanoma. We also discussed the requirements for pre-analytical and analytical as well as clinical validation process as applied to biomarkers for cancer immunotherapy. The concept of the fit-for-purpose marker validation has been introduced to address the challenges and strategies for analytical and clinical validation design for specific assays.

Modulation of Sox10, HIF-1α, Survivin, and YAP by Minocycline in the Treatment of Neurodevelopmental Handicaps following Hypoxic Insult.

Premature infants are at an increased risk of developing cognitive and motor handicaps due to chronic hypoxia. Although the current therapies have reduced the incidence of these handicaps, untoward side effects abound. Using a murine model of sublethal hypoxia, we demonstrated reduction in several transcription factors that modulate expression of genes known to be involved in several neural functions. We demonstrate the induction of these genes by minocycline, a tetracycline antibiotic with noncanonical functions, in both in vitro and in vivo studies. Specifically, there was induction of genes, including Sox10, Hif1a, Hif2a, Birc5, Yap1, Epo, Bdnf, Notch1 (cleaved), Pcna, Mag, Mobp, Plp1, synapsin, Adgra2, Pecam1, and reduction in activation of caspase 3, all known to affect proliferation, apoptosis, synaptic transmission, and nerve transmission. Minocycline treatment of mouse pups reared under sublethal hypoxic conditions resulted in improvement in open field testing parameters. These studies demonstrate beneficial effects of minocycline treatment following hypoxic insult, document up-regulation of several genes associated with improved cognitive function, and support the possibility of minocycline as a potential therapeutic target in the treatment of neurodevelopmental handicaps observed in the very premature newborn population. Additionally, these studies may aid in further interpretation of the effects of minocycline in the treatment trials and animal model studies of fragile X syndrome and multiple sclerosis.

Type II p21-activated kinases (PAKs) are regulated by an autoinhibitory pseudosubstrate.

The type II p21-activated kinases (PAKs) are key effectors of RHO-family GTPases involved in cell motility, survival, and proliferation. Using a structure-guided approach, we discovered that type II PAKs are regulated by an N-terminal autoinhibitory pseudosubstrate motif centered on a critical proline residue, and that this regulation occurs independently of activation loop phosphorylation. We determined six X-ray crystal structures of either full-length PAK4 or its catalytic domain, that demonstrate the molecular basis for pseudosubstrate binding to the active state with phosphorylated activation loop. We show that full-length PAK4 is constitutively autoinhibited, but mutation of the pseudosubstrate releases this inhibition and causes increased phosphorylation of the apoptotic regulation protein Bcl-2/Bcl-X(L) antagonist causing cell death and cellular morphological changes. We also find that PAK6 is regulated by the pseudosubstrate region, indicating a common type II PAK autoregulatory mechanism. Finally, we find Src SH3, but not ß-PIX SH3, can activate PAK4. We provide a unique understanding for type II PAK regulation.

Explainable deep learning for disease activity prediction in chronic inflammatory joint diseases.

Analysing complex diseases such as chronic inflammatory joint diseases (CIJDs), where many factors influence the disease evolution over time, is a challenging task. CIJDs are rheumatic diseases that cause the immune system to attack healthy organs, mainly the joints. Different environmental, genetic and demographic factors affect disease development and progression. The Swiss Clinical Quality Management in Rheumatic Diseases (SCQM) Foundation maintains a national database of CIJDs documenting the disease management over time for 19'267 patients. We propose the Disease Activity Score Network (DAS-Net), an explainable multi-task learning model trained on patients' data with different arthritis subtypes, transforming longitudinal patient journeys into comparable representations and predicting multiple disease activity scores. First, we built a modular model composed of feed-forward neural networks, long short-term memory networks and attention layers to process the heterogeneous patient histories and predict future disease activity. Second, we investigated the utility of the model's computed patient representations (latent embeddings) to identify patients with similar disease progression. Third, we enhanced the explainability of our model by analysing the impact of different patient characteristics on disease progression and contrasted our model outcomes with medical expert knowledge. To this end, we explored multiple feature attribution methods including SHAP, attention attribution and feature weighting using case-based similarity. Our model outperforms temporal and non-temporal neural network, tree-based, and naive static baselines in predicting future disease activity scores. To identify similar patients, a k-nearest neighbours regression algorithm applied to the model's computed latent representations outperforms baseline strategies that use raw input features representation.

Longitudinal cell-free DNA characterization by low-coverage whole-genome sequencing in patients undergoing high-dose radiotherapy.

BACKGROUND AND PURPOSE: Current radiotherapy guidelines rely heavily on imaging-based monitoring. Liquid biopsy monitoring promises to complement imaging by providing frequent systemic information about the tumor. In particular, cell-free DNA (cfDNA) sequencing offers a tumor-agnostic approach, which lends itself to monitoring heterogeneous cohorts of cancer patients. METHODS: We collected plasma cfDNA from oligometastatic patients (OMD) and head-and-neck cancer patients (SCCHN) at six time points before, during, and after radiotherapy, and compared them to the plasma samples of healthy and polymetastatic volunteers. We performed low-pass (on average 7x) whole-genome sequencing on 93 plasma cfDNA samples and correlated copy number alterations and fragment length distributions to clinical and imaging findings. RESULTS: We observed copy number alterations in 4/7 polymetastatic cancer patients, 1/7 OMD and 1/7 SCCHN patients, these patients' imaging showed progression following radiotherapy. Using unsupervised learning, we identified cancer-specific fragment length features that showed a strong correlation with copy number-based tumor fraction estimates. In 4/4 HPV-positive SCCHN patient samples, we detected viral DNA that enabled the monitoring of very low tumor fraction samples. CONCLUSIONS: Our results indicate that an elevated tumor fraction is associated with tumor aggressiveness and systemic tumor spread. This information may be used to adapt treatment strategies. Further, we show that by detecting specific sequences such as viral DNA, the sensitivity of detecting cancer from cell-free DNA sequencing data can be greatly increased.