Prediction of molecular subclasses of uveal melanoma by deep learning using routine haematoxylin-eosin-stained tissue slides.

AIMS: Uveal melanoma has a high propensity to metastasize. Prognosis is associated with specific driver mutations and copy number variations, and these can only be obtained after genetic testing. In this study we evaluated the efficacy of patient outcome prediction using deep learning on haematoxylin and eosin (HE)-stained primary uveal melanoma slides in comparison to molecular testing. METHODS: In this retrospective study of patients with uveal melanoma, 113 patients from the Erasmus Medical Centre who underwent enucleation had tumour tissue analysed for molecular classification between 1993 and 2020. Routine HE-stained slides were scanned to obtain whole-slide images (WSI). After annotation of regions of interest, tiles of 1024 × 1024 pixels were extracted at a magnification of 40×. An ablation study to select the best-performing deep-learning model was carried out using three state-of-the-art deep-learning models (EfficientNet, Vision Transformer, and Swin Transformer). RESULTS: Deep-learning models were subjected to a training cohort (n = 40), followed by a validation cohort (n = 20), and finally underwent a test cohort (n = 48). A k-fold cross-validation (k = 3) of validation and test cohorts (n = 113 of three classes: BAP1, SF3B1, EIF1AX) demonstrated Swin Transformer as the best-performing deep-learning model to predict molecular subclasses based on HE stains. The model achieved an accuracy of 0.83 ± 0.09 on the validation cohort and 0.75 ± 0.04 on the test cohort. Within the subclasses, this model correctly predicted 70% BAP1-mutated, 61% SF3B1-mutated and 80% EIF1AX-mutated UM in the test set. CONCLUSIONS: This study showcases the potential of the deep-learning methodology for predicting molecular subclasses in a multiclass manner using HE-stained WSI. This development holds promise for advanced prognostication of UM patients without the need of molecular or immunohistochemical testing. Additionally, this study suggests there are distinct histopathological features per subclass; mainly utilizing epithelioid cellular morphology for BAP1-classification, but an unknown feature distinguishes EIF1AX and SF3B1.

Cherry on Top or Real Need? A Review of Explainable Machine Learning in Kidney Transplantation.

Research on solid organ transplantation has taken advantage of the substantial acquisition of medical data and the use of artificial intelligence (AI) and machine learning (ML) to answer diagnostic, prognostic, and therapeutic questions for many years. Nevertheless, despite the question of whether AI models add value to traditional modeling approaches, such as regression models, their "black box" nature is one of the factors that have hindered the translation from research to clinical practice. Several techniques that make such models understandable to humans were developed with the promise of increasing transparency in the support of medical decision-making. These techniques should help AI to close the gap between theory and practice by yielding trust in the model by doctors and patients, allowing model auditing, and facilitating compliance with emergent AI regulations. But is this also happening in the field of kidney transplantation? This review reports the use and explanation of "black box" models to diagnose and predict kidney allograft rejection, delayed graft function, graft failure, and other related outcomes after kidney transplantation. In particular, we emphasize the discussion on the need (or not) to explain ML models for biological discovery and clinical implementation in kidney transplantation. We also discuss promising future research paths for these computational tools.

The heterogeneous sensitivity of pediatric brain tumors to different oncolytic viruses is predicted by unique gene expression profiles.

Despite decades of research, the prognosis of high-grade pediatric brain tumors (PBTs) remains dismal; however, recent cases of favorable clinical responses were documented in clinical trials using oncolytic viruses (OVs). In the current study, we employed four different species of OVs: adenovirus Delta24-RGD, herpes simplex virus rQNestin34.5v1, reovirus R124, and the non-virulent Newcastle disease virus rNDV-F0-GFP against three entities of PBTs (high-grade gliomas, atypical teratoid/rhabdoid tumors, and ependymomas) to determine their in vitro efficacy. These four OVs were screened on 14 patient-derived PBT cell cultures and the degree of oncolysis was assessed using an ATP-based assay. Subsequently, the observed viral efficacies were correlated to whole transcriptome data and Gene Ontology analysis was performed. Although no significant tumor type-specific OV efficacy was observed, the analysis revealed the intrinsic biological processes that associated with OV efficacy. The predictive power of the identified expression profiles was further validated in vitro by screening additional PBTs. In summary, our results demonstrate OV susceptibility of multiple patient-derived PBT entities and the ability to predict in vitro responses to OVs using unique expression profiles. Such profiles may hold promise for future OV preselection with effective oncolytic potency in a specific tumor, therewith potentially improving OV responses.

FAIR data retrieval for sensitive clinical research data in Galaxy.

BACKGROUND: In clinical research, data have to be accessible and reproducible, but the generated data are becoming larger and analysis complex. Here we propose a platform for Findable, Accessible, Interoperable, and Reusable (FAIR) data access and creating reproducible findings. Standardized access to a major genomic repository, the European Genome-Phenome Archive (EGA), has been achieved with API services like PyEGA3. We aim to provide a FAIR data analysis service in Galaxy by retrieving genomic data from the EGA and provide a generalized "omics" platform for FAIR data analysis. RESULTS: To demonstrate this, we implemented an end-to-end Galaxy workflow to replicate the findings from an RD-Connect synthetic dataset Beyond the 1 Million Genomes (synB1MG) available from the EGA. We developed the PyEGA3 connector within Galaxy to easily download multiple datasets from the EGA. We added the gene.iobio tool, a diagnostic environment for precision genomics, to Galaxy and demonstrate that it provides a more dynamic and interpretable view for trio analysis results. We developed a Galaxy trio analysis workflow to determine the pathogenic variants from the synB1MG trios using the GEMINI and gene.iobio tool. The complete workflow is available at WorkflowHub, and an associated tutorial was created in the Galaxy Training Network, which helps researchers unfamiliar with Galaxy to run the workflow. CONCLUSIONS: We showed the feasibility of reusing data from the EGA in Galaxy via PyEGA3 and validated the workflow by rediscovering spiked-in variants in synthetic data. Finally, we improved existing tools in Galaxy and created a workflow for trio analysis to demonstrate the value of FAIR genomics analysis in Galaxy.

Multiomic profiling of transplant glomerulopathy reveals a novel T-cell dominant subclass.

Kidney transplant (KTx) biopsies showing transplant glomerulopathy (TG) (glomerular basement membrane double contours (cg) > 0) and microvascular inflammation (MVI) in the absence of C4d staining and donor-specific antibodies (DSAs) do not fulfill the criteria for chronic active antibody-mediated rejection (CA-AMR) diagnosis and do not fit into any other Banff category. To investigate this, we initiated a multicenter intercontinental study encompassing 36 cases, comparing the immunomic and transcriptomic profiles of 14 KTx biopsies classified as cg+MVI DSA-/C4d- with 22 classified as CA-AMR DSA+/C4d+ through novel transcriptomic analysis using the NanoString Banff-Human Organ Transplant (B-HOT) panel and subsequent orthogonal subset analysis using two innovative 5-marker multiplex immunofluorescent panels. Nineteen genes were differentially expressed between the two study groups. Samples diagnosed with CA-AMR DSA+/C4d+ showed a higher glomerular abundance of natural killer cells and higher transcriptomic cell type scores for macrophages in an environment characterized by increased expression of complement-related genes (i.e., C5AR1) and higher activity of angiogenesis, interstitial fibrosis tubular atrophy, CA-AMR, and DSA-related pathways when compared to samples diagnosed with cg+MVI DSA-/C4d-. Samples diagnosed with cg+MVI DSA-/C4d- displayed a higher glomerular abundance and activity of T cells (CD3+, CD3+CD8+, and CD3+CD8-). Thus, we show that using novel multiomic techniques, KTx biopsies with cg+MVI DSA-/C4d- have a prominent T-cell presence and activity, putting forward the possibility that these represent a more T-cell dominant phenotype.

Artificial Intelligence in Digital Pathology for Bladder Cancer: Hype or Hope? A Systematic Review.

Bladder cancer (BC) diagnosis and prediction of prognosis are hindered by subjective pathological evaluation, which may cause misdiagnosis and under-/over-treatment. Computational pathology (CPATH) can identify clinical outcome predictors, offering an objective approach to improve prognosis. However, a systematic review of CPATH in BC literature is lacking. Therefore, we present a comprehensive overview of studies that used CPATH in BC, analyzing 33 out of 2285 identified studies. Most studies analyzed regions of interest to distinguish normal versus tumor tissue and identify tumor grade/stage and tissue types (e.g., urothelium, stroma, and muscle). The cell's nuclear area, shape irregularity, and roundness were the most promising markers to predict recurrence and survival based on selected regions of interest, with >80% accuracy. CPATH identified molecular subtypes by detecting features, e.g., papillary structures, hyperchromatic, and pleomorphic nuclei. Combining clinicopathological and image-derived features improved recurrence and survival prediction. However, due to the lack of outcome interpretability and independent test datasets, robustness and clinical applicability could not be ensured. The current literature demonstrates that CPATH holds the potential to improve BC diagnosis and prediction of prognosis. However, more robust, interpretable, accurate models and larger datasets-representative of clinical scenarios-are needed to address artificial intelligence's reliability, robustness, and black box challenge.

Ex vivo drug sensitivity screening predicts response to temozolomide in glioblastoma patients and identifies candidate biomarkers.

BACKGROUND: Patient-derived glioma stem-like cells (GSCs) have become the gold-standard in neuro-oncological research; however, it remains to be established whether loss of in situ microenvironment affects the clinically-predictive value of this model. We implemented a GSC monolayer system to investigate in situ-in vitro molecular correspondence and the relationship between in vitro and patient response to temozolomide (TMZ). METHODS: DNA/RNA-sequencing was performed on 56 glioblastoma tissues and 19 derived GSC cultures. Sensitivity to TMZ was screened across 66 GSC cultures. Viability readouts were related to clinical parameters of corresponding patients and whole-transcriptome data. RESULTS: Tumour DNA and RNA sequences revealed strong similarity to corresponding GSCs despite loss of neuronal and immune interactions. In vitro TMZ screening yielded three response categories which significantly correlated with patient survival, therewith providing more specific prediction than the binary MGMT marker. Transcriptome analysis identified 121 genes related to TMZ sensitivity of which 21were validated in external datasets. CONCLUSION: GSCs retain patient-unique hallmark gene expressions despite loss of their natural environment. Drug screening using GSCs predicted patient response to TMZ more specifically than MGMT status, while transcriptome analysis identified potential biomarkers for this response. GSC drug screening therefore provides a tool to improve drug development and precision medicine for glioblastoma.

Feasibility and Potential of Transcriptomic Analysis Using the NanoString nCounter Technology to Aid the Classification of Rejection in Kidney Transplant Biopsies.

BACKGROUND: Transcriptome analysis could be an additional diagnostic parameter in diagnosing kidney transplant (KTx) rejection. Here, we assessed feasibility and potential of NanoString nCounter analysis of KTx biopsies to aid the classification of rejection in clinical practice using both the Banff-Human Organ Transplant (B-HOT) panel and a customized antibody-mediated rejection (AMR)-specific NanoString nCounter Elements (Elements) panel. Additionally, we explored the potential for the classification of KTx rejection building and testing a classifier within our dataset. METHODS: Ninety-six formalin-fixed paraffin-embedded KTx biopsies were retrieved from the archives of the ErasmusMC Rotterdam and the University Hospital Cologne. Biopsies with AMR, borderline or T cell-mediated rejections (BLorTCMR), and no rejection were compared using the B-HOT and Elements panels. RESULTS: High correlation between gene expression levels was found when comparing the 2 chemistries pairwise (r = 0.76-0.88). Differential gene expression (false discovery rate; P < 0.05) was identified in biopsies diagnosed with AMR (B-HOT: 294; Elements: 76) and BLorTCMR (B-HOT: 353; Elements: 57) compared with no rejection. Using the most predictive genes from the B-HOT analysis and the Element analysis, 2 least absolute shrinkage and selection operators-based regression models to classify biopsies as AMR versus no AMR (BLorTCMR or no rejection) were developed achieving an receiver-operating-characteristic curve of 0.994 and 0.894, sensitivity of 0.821 and 0.480, and specificity of 1.00 and 0.979, respectively, during cross-validation. CONCLUSIONS: Transcriptomic analysis is feasible on KTx biopsies previously used for diagnostic purposes. The B-HOT panel has the potential to differentiate AMR from BLorTCMR or no rejection and could prove valuable in aiding kidney transplant rejection classification.

Immunomodulatory Effects of Stereotactic Body Radiotherapy and Vaccination with Heat-Killed Mycobacterium Obuense (IMM-101) in Patients with Locally Advanced Pancreatic Cancer.

BACKGROUND: Patients with locally advanced pancreatic cancer (LAPC) are treated with chemotherapy. In selected cases, stereotactic body radiotherapy (SBRT) can be added to the regimen. We hypothesized that adding an adjuvant containing a heat-killed mycobacterium (IMM-101) to SBRT may lead to beneficial immuno-modulatory effects, thereby improving survival. This study aims to investigate the safety of adding IMM-101 to SBRT and to investigate the immuno-modulatory effects of the combination treatment in the peripheral blood of LAPC patients. METHODS: LAPC patients were treated with SBRT (40 Gy) and six intradermal vaccinations of one milligram IMM-101. The primary endpoint was an observed toxicity rate of grade 4 or higher. Targeted gene-expression profiling and multicolor flow cytometry were performed for longitudinal immune-monitoring of the peripheral blood. RESULTS: Twenty patients received study treatment. No treatment-related adverse events of grade 4 or higher occurred. SBRT/IMM-101 treatment induced a transient decrease in different lymphocyte subsets and an increase in CD14+CD16-CD11b+HLA-DRlow myeloid-derived suppressor cells. Importantly, treatment significantly increased activated ICOS+, HLA-DR+ and Ki67+PD1+ T and NK cell frequencies. This was not accompanied by increased levels of most inhibitory markers, such as TIM-3 and LAG-3. CONCLUSIONS: Combination therapy with SBRT and a heat-killed mycobacterium vaccine was safe and had an immune-stimulatory effect.

Deciphering imprints of impaired memory B-cell maturation in germinal centers of three patients with common variable immunodeficiency.

Common variable immunodeficiency (CVID), characterized by recurrent infections, low serum class-switched immunoglobulin isotypes, and poor antigen-specific antibody responses, comprises a heterogeneous patient population in terms of clinical presentation and underlying etiology. The diagnosis is regularly associated with a severe decrease of germinal center (GC)-derived B-cell populations in peripheral blood. However, data from B-cell differentiation within GC is limited. We present a multiplex approach combining histology, flow cytometry, and B-cell receptor repertoire analysis of sorted GC B-cell populations allowing the modeling of distinct disturbances in GCs of three CVID patients. Our results reflect pathophysiological heterogeneity underlying the reduced circulating pool of post-GC memory B cells and plasmablasts in the three patients. In patient 1, quantitative and qualitative B-cell development in GCs is relatively normal. In patient 2, irregularly shaped GCs are associated with reduced somatic hypermutation (SHM), antigen selection, and class-switching, while in patient 3, high SHM, impaired antigen selection, and class-switching with large single clones imply increased re-cycling of cells within the irregularly shaped GCs. In the lymph nodes of patients 2 and 3, only limited numbers of memory B cells and plasma cells are formed. While reduced numbers of circulating post GC B cells are a general phenomenon in CVID, the integrated approach exemplified distinct defects during GC maturation ranging from near normal morphology and function to severe disturbances with different facets of impaired maturation of memory B cells and/or plasma cells. Integrated dissection of disturbed GC B-cell maturation by histology, flow cytometry, and BCR repertoire analysis contributes to unraveling defects in the essential steps during memory formation.

A Decentralized Kidney Transplant Biopsy Classifier for Transplant Rejection Developed Using Genes of the Banff-Human Organ Transplant Panel.

Introduction: A decentralized and multi-platform-compatible molecular diagnostic tool for kidney transplant biopsies could improve the dissemination and exploitation of this technology, increasing its clinical impact. As a first step towards this molecular diagnostic tool, we developed and validated a classifier using the genes of the Banff-Human Organ Transplant (B-HOT) panel extracted from a historical Molecular Microscope® Diagnostic system microarray dataset. Furthermore, we evaluated the discriminative power of the B-HOT panel in a clinical scenario. Materials and Methods: Gene expression data from 1,181 kidney transplant biopsies were used as training data for three random forest models to predict kidney transplant biopsy Banff categories, including non-rejection (NR), antibody-mediated rejection (ABMR), and T-cell-mediated rejection (TCMR). Performance was evaluated using nested cross-validation. The three models used different sets of input features: the first model (B-HOT Model) was trained on only the genes included in the B-HOT panel, the second model (Feature Selection Model) was based on sequential forward feature selection from all available genes, and the third model (B-HOT+ Model) was based on the combination of the two models, i.e. B-HOT panel genes plus highly predictive genes from the sequential forward feature selection. After performance assessment on cross-validation, the best-performing model was validated on an external independent dataset based on a different microarray version. Results: The best performances were achieved by the B-HOT+ Model, a multilabel random forest model trained on B-HOT panel genes with the addition of the 6 most predictive genes of the Feature Selection Model (ST7, KLRC4-KLRK1, TRBC1, TRBV6-5, TRBV19, and ZFX), with a mean accuracy of 92.1% during cross-validation. On the validation set, the same model achieved Area Under the ROC Curve (AUC) of 0.965 and 0.982 for NR and ABMR respectively. Discussion: This kidney transplant biopsy classifier is one step closer to the development of a decentralized kidney transplant biopsy classifier that is effective on data derived from different gene expression platforms. The B-HOT panel proved to be a reliable highly-predictive panel for kidney transplant rejection classification. Furthermore, we propose to include the aforementioned 6 genes in the B-HOT panel for further optimization of this commercially available panel.

The diagnostic value of nasal microbiota and clinical parameters in a multi-parametric prediction model to differentiate bacterial versus viral infections in lower respiratory tract infections.

BACKGROUND: The ability to accurately distinguish bacterial from viral infection would help clinicians better target antimicrobial therapy during suspected lower respiratory tract infections (LRTI). Although technological developments make it feasible to rapidly generate patient-specific microbiota profiles, evidence is required to show the clinical value of using microbiota data for infection diagnosis. In this study, we investigated whether adding nasal cavity microbiota profiles to readily available clinical information could improve machine learning classifiers to distinguish bacterial from viral infection in patients with LRTI. RESULTS: Various multi-parametric Random Forests classifiers were evaluated on the clinical and microbiota data of 293 LRTI patients for their prediction accuracies to differentiate bacterial from viral infection. The most predictive variable was C-reactive protein (CRP). We observed a marginal prediction improvement when 7 most prevalent nasal microbiota genera were added to the CRP model. In contrast, adding three clinical variables, absolute neutrophil count, consolidation on X-ray, and age group to the CRP model significantly improved the prediction. The best model correctly predicted 85% of the 'bacterial' patients and 82% of the 'viral' patients using 13 clinical and 3 nasal cavity microbiota genera (Staphylococcus, Moraxella, and Streptococcus). CONCLUSIONS: We developed high-accuracy multi-parametric machine learning classifiers to differentiate bacterial from viral infections in LRTI patients of various ages. We demonstrated the predictive value of four easy-to-collect clinical variables which facilitate personalized and accurate clinical decision-making. We observed that nasal cavity microbiota correlate with the clinical variables and thus may not add significant value to diagnostic algorithms that aim to differentiate bacterial from viral infections.

FAIR Genomes metadata schema promoting Next Generation Sequencing data reuse in Dutch healthcare and research.

The genomes of thousands of individuals are profiled within Dutch healthcare and research each year. However, this valuable genomic data, associated clinical data and consent are captured in different ways and stored across many systems and organizations. This makes it difficult to discover rare disease patients, reuse data for personalized medicine and establish research cohorts based on specific parameters. FAIR Genomes aims to enable NGS data reuse by developing metadata standards for the data descriptions needed to FAIRify genomic data while also addressing ELSI issues. We developed a semantic schema of essential data elements harmonized with international FAIR initiatives. The FAIR Genomes schema v1.1 contains 110 elements in 9 modules. It reuses common ontologies such as NCIT, DUO and EDAM, only introducing new terms when necessary. The schema is represented by a YAML file that can be transformed into templates for data entry software (EDC) and programmatic interfaces (JSON, RDF) to ease genomic data sharing in research and healthcare. The schema, documentation and MOLGENIS reference implementation are available at https://fairgenomes.org .

CRISPRs in the human genome are differentially expressed between malignant and normal adjacent to tumor tissue.

Clustered Regularly Interspaced Short Palindromic Repeats (CRISPRs) have been identified in bacteria, archaea and mitochondria of plants, but not in eukaryotes. Here, we report the discovery of 12,572 putative CRISPRs randomly distributed across the human chromosomes, which we termed hCRISPRs. By using available transcriptome datasets, we demonstrate that hCRISPRs are distinctively expressed as small non-coding RNAs (sncRNAs) in cell lines and human tissues. Moreover, expression patterns thereof enabled us to distinguish normal from malignant tissues. In prostate cancer, we confirmed the differential hCRISPR expression between normal adjacent and malignant primary prostate tissue by RT-qPCR and demonstrate that the SHERLOCK and DETECTR dipstick tools are suitable to detect these sncRNAs. We anticipate that the discovery of CRISPRs in the human genome can be further exploited for diagnostic purposes in cancer and other medical conditions, which certainly will lead to the development of point-of-care tests based on the differential expression of the hCRISPRs.

Training Infrastructure as a Service.

BACKGROUND: Hands-on training, whether in bioinformatics or other domains, often requires significant technical resources and knowledge to set up and run. Instructors must have access to powerful compute infrastructure that can support resource-intensive jobs running efficiently. Often this is achieved using a private server where there is no contention for the queue. However, this places a significant prerequisite knowledge or labor barrier for instructors, who must spend time coordinating deployment and management of compute resources. Furthermore, with the increase of virtual and hybrid teaching, where learners are located in separate physical locations, it is difficult to track student progress as efficiently as during in-person courses. FINDINGS: Originally developed by Galaxy Europe and the Gallantries project, together with the Galaxy community, we have created Training Infrastructure-as-a-Service (TIaaS), aimed at providing user-friendly training infrastructure to the global training community. TIaaS provides dedicated training resources for Galaxy-based courses and events. Event organizers register their course, after which trainees are transparently placed in a private queue on the compute infrastructure, which ensures jobs complete quickly, even when the main queue is experiencing high wait times. A built-in dashboard allows instructors to monitor student progress. CONCLUSIONS: TIaaS provides a significant improvement for instructors and learners, as well as infrastructure administrators. The instructor dashboard makes remote events not only possible but also easy. Students experience continuity of learning, as all training happens on Galaxy, which they can continue to use after the event. In the past 60 months, 504 training events with over 24,000 learners have used this infrastructure for Galaxy training.

Spatial genomics reveals a high number and specific location of B cells in the pancreatic ductal adenocarcinoma microenvironment of long-term survivors.

Background and aim: Only 10% of pancreatic ductal adenocarcinoma (PDAC) patients survive longer than five years. Factors underlining long-term survivorship in PDAC are not well understood. Therefore, we aimed to identify the key players in the tumor immune microenvironment (TIME) associated with long-term survivorship in PDAC patients. Methods: The immune-related gene expression profiles of resected PDAC tumors of patients who survived and remained recurrence-free of disease for ≥36 months (long-term survivors, n=10) were compared to patients who had survived ≤6 months (short-term survivors, n=10) due to tumor recurrence. Validation was performed by the spatial protein expression profile of immune cells using the GeoMx™ Digital Spatial Profiler. An independent cohort of samples consisting of 12 long-term survivors and 10 short-term survivors, was used for additional validation. The independent validation was performed by combining qualitative immunohistochemistry and quantitative protein expression profiling. Results: B cells were found to be significantly increased in the TIME of long-term survivors by gene expression profiling (p=0.018). The high tumor infiltration of B cells was confirmed by spatial protein profiling in the discovery and the validation cohorts (p=0.002 and p=0.01, respectively). The higher number of infiltrated B cells was found mainly in the stromal compartments of PDAC samples and was exclusively found within tumor cells in long-term survivors. Conclusion: This is the first comprehensive study that connects the immune landscape of gene expression profiles and protein spatial infiltration with the survivorship of PDAC patients. We found a higher number and a specific location of B cells in TIME of long-term survivors which emphasizes the importance of B cells and B cell-based therapy for future personalized immunotherapy in PDAC patients.

Deciphering Tumour Microenvironment of Liver Cancer through Deconvolution of Bulk RNA-Seq Data with Single-Cell Atlas.

Liver cancers give rise to a heavy burden on healthcare worldwide. Understanding the tumour microenvironment (TME) underpins the development of precision therapy. Single-cell RNA sequencing (scRNA-seq) technology has generated high-quality cell atlases of the TME, but its wider application faces enormous costs for various clinical circumstances. Fortunately, a variety of deconvolution algorithms can instead repurpose bulk RNA-seq data, alleviating the need for generating scRNA-seq datasets. In this study, we reviewed major public omics databases for relevance in this study and utilised eight RNA-seqs and one microarray dataset from clinical studies. To decipher the TME of liver cancer, we estimated the fractions of liver cell components by deconvoluting the samples with Cibersortx using three reference scRNA-seq atlases. We also confirmed that Cibersortx can accurately deconvolute cell types/subtypes of interest. Compared with non-tumorous liver, liver cancers showed multiple decreased cell types forming normal liver microarchitecture, as well as elevated cell types involved in fibrogenesis, abnormal angiogenesis, and disturbed immune responses. Survival analysis shows that the fractions of five cell types/subtypes significantly correlated with patient outcomes, indicating potential therapeutic targets. Therefore, deconvolution of bulk RNA-seq data with scRNA-seq atlas references can be a useful tool to help understand the TME.

Robust deep learning model for prognostic stratification of pancreatic ductal adenocarcinoma patients.

A major challenge for treating patients with pancreatic ductal adenocarcinoma (PDAC) is the unpredictability of their prognoses due to high heterogeneity. We present Multi-Omics DEep Learning for Prognosis-correlated subtyping (MODEL-P) to identify PDAC subtypes and to predict prognoses of new patients. MODEL-P was trained on autoencoder integrated multi-omics of 146 patients with PDAC together with their survival outcome. Using MODEL-P, we identified two PDAC subtypes with distinct survival outcomes (median survival 10.1 and 22.7 months, respectively, log rank p = 1 × 10-6), which correspond to DNA damage repair and immune response. We rigorously validated MODEL-P by stratifying patients in five independent datasets into these two survival groups and achieved significant survival difference, which is superior to current practice and other subtyping schemas. We believe the subtype-specific signatures would facilitate PDAC pathogenesis discovery, and MODEL-P can provide clinicians the prognoses information in the treatment decision-making to better gauge the benefits versus the risks.

Fusion transcripts and their genomic breakpoints in polyadenylated and ribosomal RNA-minus RNA sequencing data.

BACKGROUND: Fusion genes are typically identified by RNA sequencing (RNA-seq) without elucidating the causal genomic breakpoints. However, non-poly(A)-enriched RNA-seq contains large proportions of intronic reads that also span genomic breakpoints. RESULTS: We have developed an algorithm, Dr. Disco, that searches for fusion transcripts by taking an entire reference genome into account as search space. This includes exons but also introns, intergenic regions, and sequences that do not meet splice junction motifs. Using 1,275 RNA-seq samples, we investigated to what extent genomic breakpoints can be extracted from RNA-seq data and their implications regarding poly(A)-enriched and ribosomal RNA-minus RNA-seq data. Comparison with whole-genome sequencing data revealed that most genomic breakpoints are not, or minimally, transcribed while, in contrast, the genomic breakpoints of all 32 TMPRSS2-ERG-positive tumours were present at RNA level. We also revealed tumours in which the ERG breakpoint was located before ERG, which co-existed with additional deletions and messenger RNA that incorporated intergenic cryptic exons. In breast cancer we identified rearrangement hot spots near CCND1 and in glioma near CDK4 and MDM2 and could directly associate this with increased expression. Furthermore, in all datasets we find fusions to intergenic regions, often spanning multiple cryptic exons that potentially encode neo-antigens. Thus, fusion transcripts other than classical gene-to-gene fusions are prominently present and can be identified using RNA-seq. CONCLUSION: By using the full potential of non-poly(A)-enriched RNA-seq data, sophisticated analysis can reliably identify expressed genomic breakpoints and their transcriptional effects.

OralDisk: A Chair-Side Compatible Molecular Platform Using Whole Saliva for Monitoring Oral Health at the Dental Practice.

Periodontitis and dental caries are two major bacterially induced, non-communicable diseases that cause the deterioration of oral health, with implications in patients' general health. Early, precise diagnosis and personalized monitoring are essential for the efficient prevention and management of these diseases. Here, we present a disk-shaped microfluidic platform (OralDisk) compatible with chair-side use that enables analysis of non-invasively collected whole saliva samples and molecular-based detection of ten bacteria: seven periodontitis-associated (Aggregatibacter actinomycetemcomitans, Campylobacter rectus, Fusobacterium nucleatum, Prevotella intermedia, Porphyromonas gingivalis, Tannerella forsythia, Treponema denticola) and three caries-associated (oral Lactobacilli, Streptococcus mutans, Streptococcus sobrinus). Each OralDisk test required 400 µL of homogenized whole saliva. The automated workflow included bacterial DNA extraction, purification and hydrolysis probe real-time PCR detection of the target pathogens. All reagents were pre-stored within the disk and sample-to-answer processing took < 3 h using a compact, customized processing device. A technical feasibility study (25 OralDisks) was conducted using samples from healthy, periodontitis and caries patients. The comparison of the OralDisk with a lab-based reference method revealed a ~90% agreement amongst targets detected as positive and negative. This shows the OralDisk's potential and suitability for inclusion in larger prospective implementation studies in dental care settings.