Deep learning integrates histopathology and proteogenomics at a pan-cancer level

Pan-cancer analysis shapes the understanding of cancer biology and medicine

Advances in multi-omics datasets and analytical methods have revolutionized cancer research, offering a comprehensive, pan-cancer perspective. Pan-cancer studies identify shared mechanisms and unique traits across different cancer types, which are reshaping diagnostic and treatment strategies. However, continued innovation is required to refine these approaches and deepen our understanding of cancer biology and medicine. This review summarized key findings from pan-cancer research and  explored their potential to drive future advancements in oncology.

Artificial intelligence-enabled automated analysis of transmission electron micrographs to evaluate chemotherapy impact on mitochondrial morphology in triple negative breast cancer

Advancements in transmission electron microscopy (TEM) have enabled in-depth studies of biological specimens, offering new avenues to large-scale imaging experiments with subcellular resolution. Mitochondrial structure is of growing interest in cancer biology due to its crucial role in regulating the multi-faceted functions of mitochondria. We and others have established the crucial role of mitochondria in triple-negative breast cancer (TNBC), an aggressive subtype of breast cancer with limited therapeutic options. Building upon our previous work demonstrating the regulatory role of mitochondrial structure dynamics in metabolic adaptation and survival of chemotherapy-refractory TNBC cells, we sought to extend those findings to a large-scale analysis of transmission electron micrographs. Here we present a UNet artificial intelligence (AI) model for automatic annotation and assessment of mitochondrial morphology and feature quantification. Our model is trained on 11,039 manually annotated mitochondria across 125 micrographs derived from a variety of orthotopic patient-derived xenograft (PDX) mouse model tumors and adherent cell cultures. The model achieves an F1 score of 0.85 on test micrographs at the pixel level. To validate the ability of our model to detect expected mitochondrial structural features, we utilized micrographs from mouse primary skeletal muscle cells genetically modified to lack Dynamin-related protein 1 (Drp1). The algorithm successfully detected a significant increase in mitochondrial elongation, in alignment with the well-established role of Drp1 as a driver of mitochondrial fission. Further, we subjected in vitro and in vivo TNBC models to conventional chemotherapy treatments commonly used for clinical management of TNBC, including doxorubicin, carboplatin, paclitaxel, and docetaxel (DTX). We found substantial within-sample heterogeneity of mitochondrial structure in both in vitro and in vivo TNBC models and observed a consistent reduction in mitochondrial elongation in DTX-treated specimens. We went on to compare mammary tumors and matched lung metastases in a highly metastatic PDX model of TNBC, uncovering significant reduction in mitochondrial length in metastatic lesions. Our large, curated dataset provides high statistical power to detect frequent chemotherapy-induced shifts in mitochondrial shapes and sizes in residual cells left behind after treatment. The successful application of our AI model to capture mitochondrial structure marks a step forward in high-throughput analysis of mitochondrial structures, enhancing our understanding of how morphological changes may relate to chemotherapy efficacy and mechanism of action. Finally, our large, manually curated electron micrograph dataset - now publicly available - serves as a unique gold-standard resource for developing, benchmarking, and applying computational models, while further advancing investigations into mitochondrial morphology and its impact on cancer biology.

Deep learning-based multi-task prediction of response to neoadjuvant chemotherapy using multiscale whole slide images in breast cancer: A multicenter study

Objective

Early predicting response before neoadjuvant chemotherapy (NAC) is crucial for personalized treatment plans for locally advanced breast cancer patients. We aim to develop a multi-task model using multiscale whole slide images (WSIs) features to predict the response to breast cancer NAC more finely.

Methods

This work collected 1,670 whole slide images for training and validation sets, internal testing sets, external testing sets, and prospective testing sets of the weakly-supervised deep learning-based multi-task model (DLMM) in predicting treatment response and pCR to NAC. Our approach models two-by-two feature interactions across scales by employing concatenate fusion of single-scale feature representations, and controls the expressiveness of each representation via a gating-based attention mechanism.

Results

In the retrospective analysis, DLMM exhibited excellent predictive performance for the prediction of treatment response, with area under the receiver operating characteristic curves (AUCs) of 0.869 [95% confidence interval (95% CI): 0.806-0.933] in the internal testing set and 0.841 (95% CI: 0.814-0.867) in the external testing sets. For the pCR prediction task, DLMM reached AUCs of 0.865 (95% CI: 0.763-0.964) in the internal testing and 0.821 (95% CI: 0.763-0.878) in the pooled external testing set. In the prospective testing study, DLMM also demonstrated favorable predictive performance, with AUCs of 0.829 (95% CI: 0.754-0.903) and 0.821 (95% CI: 0.692-0.949) in treatment response and pCR prediction, respectively. DLMM significantly outperformed the baseline models in all testing sets (P<0.05). Heatmaps were employed to interpret the decision-making basis of the model. Furthermore, it was discovered that high DLMM scores were associated with immune-related pathways and cells in the microenvironment during biological basis exploration.

Conclusions

The DLMM represents a valuable tool that aids clinicians in selecting personalized treatment strategies for breast cancer patients.

The 2024 Report on the Human Proteome from the HUPO Human Proteome Project

The Human Proteome Project (HPP), the flagship initiative of the Human Proteome Organization (HUPO), has pursued two goals: (1) to credibly identify at least one isoform of every protein-coding gene and (2) to make proteomics an integral part of multiomics studies of human health and disease. The past year has seen major transitions for the HPP. neXtProt was retired as the official HPP knowledge base, UniProtKB became the reference proteome knowledge base, and Ensembl-GENCODE provides the reference protein target list. A function evidence FE1-5 scoring system has been developed for functional annotation of proteins, parallel to the PE1-5 UniProtKB/neXtProt scheme for evidence of protein expression. This report includes updates from neXtProt (version 2023-09) and UniProtKB release 2024_04, with protein expression detected (PE1) for 18138 of the 19411 GENCODE protein-coding genes (93%). The number of non-PE1 proteins ("missing proteins") is now 1273. The transition to GENCODE is a net reduction of 367 proteins (19,411 PE1-5 instead of 19,778 PE1-4 last year in neXtProt). We include reports from the Biology and Disease-driven HPP, the Human Protein Atlas, and the HPP Grand Challenge Project. We expect the new Functional Evidence FE1-5 scheme to energize the Grand Challenge Project for functional annotation of human proteins throughout the global proteomics community, including π-HuB in China.

Deep learning application in prediction of cancer molecular alterations based on pathological images: a bibliographic analysis via CiteSpace

BACKGROUND

The advancements in artificial intelligence (AI) technology for image recognition were propelling molecular pathology research into a new era.

OBJECTIVE

To summarize the hot spots and research trends in the field of molecular pathology image recognition.

METHODS

Relevant articles from January 1st, 2010, to August 25th, 2023, were retrieved from the Web of Science Core Collection. Subsequently, CiteSpace was employed for bibliometric and visual analysis, generating diverse network diagrams illustrating keywords, highly cited references, hot topics, and research trends.

RESULTS

A total of 110 relevant articles were extracted from a pool of 10,205 articles. The overall publication count exhibited a rising trend each year. The leading contributors in terms of institutions, countries, and authors were Maastricht University (11 articles), the United States (38 articles), and Kather Jacob Nicholas (9 articles), respectively. Half of the top ten research institutions, based on publication volume, were affiliated with Germany. The most frequently cited article was authored by Nicolas Coudray et al. accumulating 703 citations. The keyword "Deep learning" had the highest frequency in 2019. Notably, the highlighted keywords from 2022 to 2023 included "microsatellite instability", and there were 21 articles focusing on utilizing algorithms to recognize microsatellite instability (MSI) in colorectal cancer (CRC) pathological images.

CONCLUSION

The use of DL is expected to provide a new strategy to effectively solve the current problem of time-consuming and expensive molecular pathology detection. Therefore, further research is needed to address issues, such as data quality and standardization, model interpretability, and resource and infrastructure requirements.

ESCCPred: a machine learning model for diagnostic prediction of early esophageal squamous cell carcinoma using autoantibody profiles

BACKGROUND

Esophageal squamous cell carcinoma (ESCC) is a deadly cancer with no clinically ideal biomarkers for early diagnosis. The objective of this study was to develop and validate a user-friendly diagnostic tool for early ESCC detection.

METHODS

The study encompassed three phases: discovery, verification, and validation, comprising a total of 1309 individuals. Serum autoantibodies were profiled using the HuProtTM human proteome microarray, and autoantibody levels were measured using the enzyme-linked immunosorbent assay (ELISA). Twelve machine learning algorithms were employed to construct diagnostic models, and evaluated using the area under the receiver operating characteristic curve (AUC). The model application was facilitated through R Shiny, providing a graphical interface.

RESULTS

Thirteen autoantibodies targeting TAAs (CAST, FAM131A, GABPA, HDAC1, HDGFL1, HSF1, ISM2, PTMS, RNF219, SMARCE1, SNAP25, SRPK2, and ZPR1) were identified in the discovery phase. Subsequent verification and validation phases identified five TAAbs (anti-CAST, anti-HDAC1, anti-HSF1, anti-PTMS, and anti-ZPR1) that exhibited significant differences between ESCC and control subjects (P < 0.05). The support vector machine (SVM) model demonstrated robust performance, with AUCs of 0.86 (95% CI: 0.82-0.89) in the training set and 0.83 (95% CI: 0.78-0.88) in the test set. For early-stage ESCC, the SVM model achieved AUCs of 0.83 (95% CI: 0.79-0.88) in the training set and 0.83 (95% CI: 0.77-0.90) in the test set. Notably, promising results were observed for high-grade intraepithelial neoplasia, with an AUC of 0.87 (95% CI: 0.77-0.98). The web-based implementation of the early ESCC diagnostic tool is publicly accessible at https://litdong.shinyapps.io/ESCCPred/ .

CONCLUSION

This study provides a promising and easy-to-use diagnostic prediction model for early ESCC detection. It holds promise for improving early detection strategies and has potential implications for public health.

Mismatch repair-proficient tumor footprints in the sands of immune desert: mechanistic constraints and precision platforms

Mismatch repair proficient (MMRp) tumors of colorectal origin are one of the prevalent yet unpredictable clinical challenges. Despite earnest efforts, optimal treatment modalities have yet to emerge for this class. The poor prognosis and limited actionability of MMRp are ascribed to a low neoantigen burden and a desert-like microenvironment. This review focuses on the critical roadblocks orchestrated by an immune evasive mechanistic milieu in the context of MMRp. The low density of effector immune cells, their weak spatiotemporal underpinnings, and the high-handedness of the IL-17-TGF-β signaling are intertwined and present formidable challenges for the existing therapies. Microbiome niche decorated by Fusobacterium nucleatum alters the metabolic program to maintain an immunosuppressive state. We also highlight the evolving strategies to repolarize and reinvigorate this microenvironment. Reconstruction of anti-tumor chemokine signaling, rational drug combinations eliciting T cell activation, and reprograming the maladapted microbiome are exciting developments in this direction. Alternative vulnerability of other DNA damage repair pathways is gaining momentum. Integration of liquid biopsy and ex vivo functional platforms provide precision oncology insights. We illustrated the perspectives and changing landscape of MMRp-CRC. The emerging opportunities discussed in this review can turn the tide in favor of fighting the treatment dilemma for this elusive cancer.

Rapid multiplex immunohistochemistry for characterizing tumor-immune microenvironment

Intratumoral immune profiles are related to prognosis and therapeutic efficacy, and could result in personalized treatments based on biomarkers. To develop a multiplex, quantitative, and rapid tissue evaluation method based on the clinically established standard immunohistochemistry (IHC), a 6-marker rapid multiplex IHC was developed based on our previously reported 14-marker multiplex IHC by reducing the number of labels and accelerating the staining procedure. First, fewer labels were required to identify the same immunological features linked to prognosis in 14-marker multiplex IHC analyses. The six selected markers showed a significant correlation with the 14 markers in the immune classification. Next, a rapid staining protocol was developed by optimizing the reaction temperature, chromogen, and washing time, allowing the completion of 6-marker analysis in 5 h and 49 min, as opposed to the several days required for conventional multiplex IHC. Validation of benign tonsil and head and neck cancer tissues revealed a significant correlation between rapid and conventional 6-makrer multiplex IHC in terms of staining intensities, densities of T cells, macrophages, lymphoid/myeloid immune cell ratios, and spatial profiles of intratumoral immune infiltrates. This method may enable quantitative assessment of the tumor-immune microenvironment on a clinically feasible time scale, which promotes the development of tissue biomarker-guided therapeutic strategies.

Pan-cancer proteogenomics characterization of tumor immunity

Despite the successes of immunotherapy in cancer treatment over recent decades, less than <10%-20% cancer cases have demonstrated durable responses from immune checkpoint blockade. To enhance the efficacy of immunotherapies, combination therapies suppressing multiple immune evasion mechanisms are increasingly contemplated. To better understand immune cell surveillance and diverse immune evasion responses in tumor tissues, we comprehensively characterized the immune landscape of more than 1,000 tumors across ten different cancers using CPTAC pan-cancer proteogenomic data. We identified seven distinct immune subtypes based on integrative learning of cell type compositions and pathway activities. We then thoroughly categorized unique genomic, epigenetic, transcriptomic, and proteomic changes associated with each subtype. Further leveraging the deep phosphoproteomic data, we studied kinase activities in different immune subtypes, which revealed potential subtype-specific therapeutic targets. Insights from this work will facilitate the development of future immunotherapy strategies and enhance precision targeting with existing agents.