Cancer is a heterogeneous disease requiring costly genetic profiling for better understanding and management. Recent advances in deep learning have enabled cost-effective predictions of genetic alterations from whole slide images (WSIs). While transformers have driven significant progress in non-medical domains, their application to WSIs lags behind due to high model complexity and limited dataset sizes. Here, we introduce SEQUOIA, a linearized transformer model that predicts cancer transcriptomic profiles from WSIs. SEQUOIA is developed using 7584 tumor samples across 16 cancer types, with its generalization capacity validated on two independent cohorts comprising 1368 tumors. Accurately predicted genes are associated with key cancer processes, including inflammatory response, cell cycles and metabolism. Further, we demonstrate the value of SEQUOIA in stratifying the risk of breast cancer recurrence and in resolving spatial gene expression at loco-regional levels. SEQUOIA hence deciphers clinically relevant information from WSIs, opening avenues for personalized cancer management.

Background: Identification of prognostic biomarkers in cancers is a crucial step to improve overall survival (OS). Although mutations in tumour protein 53 (TP53) is prevalent in astrocytoma, the prognostic effects of TP53 mutation are unclear. Methods: In this retrospective study, we sequenced TP53 exons 1 to 10 in a cohort of 102 lower-grade glioma (LGG) subtypes and determined the prognostic effects of TP53 mutation in astrocytoma and oligodendroglioma. Publicly available datasets were analysed to confirm the findings. Results: In astrocytoma, mutations in TP53 codon 273 were associated with a significantly increased OS compared to the TP53 wild-type (HR (95% CI): 0.169 (0.036–0.766), p = 0.021). Public datasets confirmed these findings. TP53 codon 273 mutant astrocytomas were significantly more chemosensitive than TP53 wild-type astrocytomas (HR (95% CI): 0.344 (0.13–0.88), p = 0.0148). Post-chemotherapy, a significant correlation between TP53 and YAP1 mRNA was found (p = 0.01). In O (6)-methylguanine methyltransferase (MGMT) unmethylated chemotherapy-treated astrocytoma, both TP53 codon 273 and YAP1 mRNA were significant prognostic markers. In oligodendroglioma, TP53 mutations were associated with significantly decreased OS. Conclusions: Based on these findings, we propose that certain TP53 mutant astrocytomas are chemosensitive through the involvement of YAP1, and we outline a potential mechanism. Thus, TP53 mutations may be key drivers of astrocytoma therapeutic efficacy and influence survival outcomes.

Lower-grade glioma (LGG) is a diffuse infiltrative tumor of the central nervous system, which lacks targeted therapy. We investigated the role of Podocan-like 1 (PODNL1) methylation in LGG clinical outcomes using the TCGA-LGG transcriptomics dataset. We identified four PODNL1 CpG sites, cg07425555, cg26969888, cg18547299, and cg24354933, which were associated with unfavorable overall survival (OS) and disease-free survival (DFS) in univariate and multivariate analysis after adjusting for age, gender, tumor-grade, and IDH1-mutation. In multivariate analysis, the OS and DFS hazard ratios ranged from 0.44 to 0.58 (p < 0.001) and 0.62 to 0.72 (p < 0.001), respectively, for the four PODNL1 CpGs. Enrichment analysis of differential gene and protein expression and analysis of 24 infiltrating immune cell types showed significantly increased infiltration in LGGs and its histological subtypes with low-methylation levels of the PODNL1 CpGs. High PODNL1 expression and low-methylation subgroups of the PODNL1 CpG sites were associated with significantly increased PD-L1, PD-1, and CTLA4 expressions. PODNL1 methylation may thus be a potential indicator of immune checkpoint blockade response, and serve as a biomarker for determining prognosis and immune subtypes in LGG.

Notch signalling pathway, particularly its ligand delta-ligand 3 (DLL3), is important in glioma, however, little is known about DLL3 regulation and prognostic effects. Immunohistochemistry on a cohort of 163 gliomas revealed DLL3 upregulation in IDH1 mutant gliomas, where it was associated with a favourable prognosis (HR[95% CI]: 0.28[0.09-0.87]; p = 0.021). We investigated the epigenetic regulation of DLL3, and identified individual CpG sites correlating with DLL3 mRNA expression, which were significant prognostic markers in LGG. In silico analysis revealed that infiltrating immune cells significantly correlated with DLL3 expression, methylation and somatic copy number alterations. The prognostic effects of DLL3 expression was significantly affected by infiltration of immune cells. RNA Sequencing of 83 LGGs and GO Term analysis of differentially expressed genes showed that low DLL3 expression was related to ciliogenesis, which was confirmed by TCGA LGG analysis. Thus, DLL3 may play an important role in the immune microenvironment and prognosis of LGGs.

Background: Tumour mutation burden (TMB) is an emerging pan‐cancer biomarker with predictive value for immune checkpoint inhibitor (ICI) outcomes, yet evidence is inconsistent due to methodological variability and cut‐off thresholds. This systematic review and meta‐analysis evaluated the impact of TMB on overall survival (OS) and progression‐free survival (PFS) across solid tumours in ICI‐treated cohorts and its predictive relevance in non‐ICI‐treated cohorts. Methods: Following PRISMA 2020 guidelines, we searched PubMed, Scopus, ScienceDirect and Cochrane for studies published between 2010 and 2024 reporting hazard ratios (HRs) and 95% confidence intervals (CIs) for OS and PFS in high‐ versus low‐TMB cohorts. High and low TMB were defined by study‐specific cut‐offs, and ultra‐high TMB was defined as the top 20% of cohort‐specific values. Study quality was assessed with the Newcastle‐Ottawa Scale; heterogeneity with I ² ; publication bias with funnel plots/Egger's test; and robustness by leave‐one‐out analysis. Results: 5278 patients across 28 studies were analysed. High TMB, defined by cohort‐specific cut‐offs, was significantly associated with improved OS and PFS, particularly in non‐small cell lung cancer (OS: HR = 0.56), selected gastrointestinal cancers (OS: HR = 0.36), and advanced/recurrent tumours (OS: HR = 0.52). Benefits were greatest in ICI‐treated patients, especially with combined anti‐PD‐L1/PD‐1 and anti‐CTLA‐4 therapy (OS: HR = 0.47; PFS: HR = 0.50). Chemotherapy‐treated cohorts also showed better outcomes, but less consistently (OS: HR = 0.60; PFS: HR = 0.55). Ultra‐high TMB had better OS than the universal 10 mut/Mb cut‐off (HR = 0.44 vs. 0.58). Non‐beneficial associations were observed in glioma and penile squamous cell carcinoma, highlighting disease‐specific variability. Sequencing platforms and cut‐off definitions remained sources of heterogeneity. Conclusion: TMB demonstrates prognostic relevance and predictive utility in a histology‐ and treatment‐context‐dependent manner, with the most consistent associations in selected ICI‐treated tumours. Associations in non‐ICI‐treated cohorts were weaker and inconsistent, indicating putative predictive value. Standardising TMB assessment and refining relevant thresholds are essential for optimising its role in precision oncology. Trial Registration: PROSPERO Registration Number: CRD42024608809

Cancer involves dynamic changes caused by (epi)genetic alterations such as mutations or abnormal DNA methylation patterns which occur in cancer driver genes. These driver genes are divided into oncogenes and tumor suppressors depending on their function and mechanism of action. Discovering driver genes in different cancer (sub)types is important not only for increasing current understanding of carcinogenesis but also from prognostic and therapeutic perspectives. We have previously developed a framework called Moonlight which uses a systems biology multi-omics approach for prediction of driver genes. Here, we present an important development in Moonlight2 by incorporating a DNA methylation layer which provides epigenetic evidence for deregulated expression profiles of driver genes. To this end, we present a novel functionality called Gene Methylation Analysis (GMA) which investigates abnormal DNA methylation patterns to predict driver genes. This is achieved by integrating the tool EpiMix which is designed to detect such aberrant DNA methylation patterns in a cohort of patients and further couples these patterns with gene expression changes. To showcase GMA, we applied it to three cancer (sub)types (basal-like breast cancer, lung adenocarcinoma, and thyroid carcinoma) where we discovered 33, 190, and 263 epigenetically driven genes, respectively. A subset of these driver genes had prognostic effects with expression levels significantly affecting survival of the patients. Moreover, a subset of the driver genes demonstrated therapeutic potential as drug targets. This study provides a framework for exploring the driving forces behind cancer and provides novel insights into the landscape of three cancer sub(types) by integrating gene expression and methylation data.