Research Areas

Chordoma is a rare, chemo-resistant bone sarcoma characterized by slow growth and a high proportion of G0 phase cells. Despite aggressive clinical recurrence, the metabolic dependencies and intrinsic resistance mechanisms against non-apoptotic cell death pathways such as ferroptosis remain elusive; bulk sequencing fails to isolate these signals due to high stromal contamination. We analyzed high-resolution scRNA-seq datasets from primary chordoma tissues and digitally isolated 9,157 tumor cells expressing the chordoma-specific diagnostic marker TBXT (Brachyury). Lipid remodeling, iron metabolism, and antioxidant defense networks were mapped at single-cell resolution with rigorous statistical thresholding to account for transcriptomic dropout. Chordoma cells exhibit profound transcriptional "metabolic lockdown": the SREBF1-ACSL4 axis driving polyunsaturated fatty acid synthesis is dormant at baseline (Pearson r=0.017), while 74.9% of TBXT⁺ cells show extreme stoichiometric dominance of GPX4 over ACSL4. The system xc⁻ antiporter is functionally silent (mean SLC7A11: 0.0067), and intracellular iron pools appear sequestered (FTH1/FTL ratio = 3.60, low IRP signature). High intercellular coefficient of variation in ER stress markers (e.g., XBP1 CV=1.24) indicates a hidden homeostatic vulnerability. Chordoma avoids lipid peroxidation through a transcriptionally static, single-point reliance on GPX4 combined with an inactive lipid-reprogramming pathway — not through dynamic multi-layered defense. These findings provide a genetic rationale that baseline chordoma is resistant to conventional ferroptosis inducers but highly vulnerable if static ER/lipid homeostasis can be disrupted.

This study presents an AI-driven approach to reclassify 9,534 germline Variants of Uncertain Significance (VUS) in ATM and PALB2 genes. By leveraging proteome-wide computational methods, we identify actionable targets for precision oncology, potentially expanding the eligible patient population for PARP inhibitor therapies. Our findings demonstrate the power of AI in resolving the clinical ambiguity of VUS classifications, transforming uncertain genetic findings into actionable therapeutic opportunities.

Computational toxicology plays a pivotal role in modern drug discovery and environmental risk assessment; however, the reliability of predictive models on unseen chemical scaffolds remains a critical bottleneck. Deep learning architectures, despite their prevalence, are susceptible to 'silent failures'—yielding high-confidence errors on out-of-distribution data—which poses significant risks in safety-critical applications. In this study, we propose a 'Safe-by-Design' hybrid framework to mitigate this epistemic uncertainty. Rather than relying on a single algorithmic paradigm, we integrate the explicit structural 'memory' of classical Random Forests with the topological 'intuition' of Graph Neural Networks (GNNs) via a transparent Stacking Ensemble (Logistic Regression) functioning as a Mixture of Experts (MoE). Evaluated across 12 Tox21 endpoints using a rigorous 5-seed benchmarking protocol under Nested Scaffold Split, our analysis demonstrates that classical models serve as an essential robustness layer in data-scarce regimes.

Rare bone tumors, including chordoma and sarcoma, present significant therapeutic challenges due to limited treatment options. This research leverages multi-omics data integration to identify novel therapeutic targets. Using advanced computational methods, we analyzed genomic, transcriptomic, and proteomic data to uncover potential druggable pathways specific to these rare malignancies.

This study investigates the intrinsic resistance mechanisms in antibody-drug conjugate (ADC) therapies for lung adenocarcinoma. Through comprehensive genomic analysis, we identified a novel 1p32 co-deletion syndrome involving TACSTD2 and CDKN2C genes, which contributes to resistance in TROP2-targeted therapies. Our findings propose a novel inhibitor as a potential rescue strategy for TROP2-negative patients.