Brain metastasis (BM) is the most common event in patients with lung cancer. Despite multimodal treatments and advances in systemic therapies, development of BM remains one of the main factors associated with poor prognosis and mortality in patients with lung cancer. Therefore, better understanding of mechanisms involved in lung cancer brain metastasis (LCBM) is of great importance to suppress cancer cells and to improve the overall survival of patients. Several cancer-related genes such as EGFR and KRAS have been proposed as potential predictors of LCBM. In addition, there is ample evidence supporting crucial roles of non-coding RNAs (ncRNAs) in mediating LCBM. In this review, we provide comprehensive information on risk assessment, predictive, and prognostic panels for early detection of BM in patients with lung cancer. Moreover, we present an overview of LCBM molecular mechanisms, cancer driver genes, and ncRNAs which may predict the risk of BM in lung cancer patients. Recent clinical studies have focused on determining mechanisms involved in LCBM and their association with diagnosis, prognosis, and treatment outcomes. These studies have shown that alterations in EGFR, KRAS, BRAF, and ALK, as the most frequent coding gene alterations, and dysregulation of ncRNAs such as miR-423, miR-330-3p, miR-145, piR-651, and MALAT1 can be considered as potential biomarkers of LCBM.

Colorectal cancer (CRC) is the third most commonly diagnosed malignancy and has the second highest mortality rate globally. Thanks to the advent of next-generation sequencing technologies, several novel candidate genes have been proposed for CRC susceptibility. Germline biallelic mutations in one or more of the 22 currently recognized Fanconi anemia (FA) genes have been associated with Fanconi anemia disease, while germline monoallelic mutations, somatic mutations, or the promoter hypermethylation of some FANC genes increases the risk of cancer development, including CRC. The FA pathway is a substantial part of the DNA damage response system that participates in the repair of DNA inter-strand crosslinks through homologous recombination (HR) and protects genome stability via replication fork stabilization, respectively. Recent studies revealed associations between FA gene/protein tumor expression levels (i.e., FANC genes) and CRC progression and drug resistance. Moreover, the FA pathway represents a potential target in the CRC treatment. In fact, FANC gene characteristics may contribute to chemosensitize tumor cells to DNA crosslinking agents such as oxaliplatin and cisplatin besides exploiting the synthetic lethal approach for selective targeting of tumor cells. Hence, this review summarizes the current knowledge on the function of the FA pathway in DNA repair and genomic integrity with a focus on the FANC genes as potential predisposition factors to CRC. We then introduce recent literature that highlights the importance of FANC genes in CRC as promising prognostic and predictive biomarkers for disease management and treatment. Finally, we represent a brief overview of the current knowledge around the FANC genes as synthetic lethal therapeutic targets for precision cancer medicine.

Synthetic lethality offers a promising approach for developing effective therapeutic interventions in cancer when direct targeting of driver genes is impractical. In this study, we comprehensively analyzed large-scale CRISPR, shRNA, and PRISM screens to identify potential synthetic lethal (SL) interactions in pan-cancer and 12 individual cancer types, using a new computational framework that leverages the biological function and signaling pathway information of key driver genes to mitigate the confounding effects of background genetic alterations in different cancer cell lines. This approach has successfully identified several putative SL interactions, includingKRAS-MAP3K2andAPC-TCF7L2in pan cancer, andCCND1-METTL1,TP53-FRS3,SMO-MDM2, andCCNE1-MTORin liver, blood, skin, and gastric cancers, respectively. In addition, we proposed several FDA-approved cancer-targeted drugs for various cancer types through PRISM drug screens, such as cabazitaxel forVHL-mutated kidney cancer and alectinib for lung cancer withNRASorKRASmutations. Leveraging pathway information can enhance the concordance of shRNA and CRISPR screens and provide clinically relevant findings such as the potential efficacy of dasatinib, an inhibitor ofSRC, for colorectal cancer patients with mutations in the WNT signaling pathway. These analyses revealed that taking signaling pathway information into account results in the identification of more promising SL interactions.

Liquid biopsies and cell-free DNA (cfDNA) offer minimally invasive methods for the diagnosis and monitoring of Ewing Sarcoma (EwS). EwS have a low tumour mutational burden and their detection with plasma cfDNA is challenging. We hypothesised that analysing the cfDNA methylome and fragmentome could enhance sensitivity for detecting EwS and identifying disease recurrence. Using T7-MBD-seq, we conducted whole-genome and methylome sequencing of cfDNA from 87 serial samples of 23 patients with EwS and 3 patients with CIC-rearranged sarcoma (CIC). With EwingSign, a new machine learning model, we identified EwS or CIC in a test set for 11 out of 16 patients at diagnosis and 15 out of 18 clinically confirmed relapse events. 0 out of 24 non-cancer controls (NCC) were detected positive with EwingSign. When combined with global and regional fragmentome analysis, all 18 relapse cases were detected, with 15/18 detected by 2 or more modalities, and 1 out of 24 NCC was detected by one modality. These findings indicate that cfDNA methylome and fragmentome analysis, if validated in a larger cohort, could improve disease detection, monitoring and relapse identification in patients with EwS.

Predicting cellular responses to genetic and chemical perturbations remains a central challenge in single-cell biology and a key step toward building in silico virtual cells. The rapid growth of perturbation datasets and advances in deep-learning models have raised expectations for accurate and generalizable prediction. We show that these expectations are overly optimistic, largely due to the failure modes of existing evaluation metrics. In this study, using cross-splitting, controlled noise experiments, and synthetic data, we systematically evaluate both prediction models and evaluation metrics. We demonstrate that widely used metrics, including correlation-based measures and common distributional distances, are strongly influenced by scale, sparsity, and dimensionality, often misrepresenting model performance. In particular, the Wasserstein distance fails in high-dimensional gene expression spaces under variance scaling, while the Energy distance can overlook disruptions in gene-gene dependencies. Our analyses further reveal that complex deep learning models often underperform simple baselines and remain far from empirical performance bounds across multiple chemical perturbation datasets. Together, our framework exposes critical pitfalls, establishes robust evaluation guidelines, and provides a foundation for trustworthy benchmarking toward reliable virtual-cell models.