Gene coordination patterns across 8,314 tumors reveal a spectral point of no return in cancer progression
Mayfield, J.
Show abstract
Genes operate through coordinated patterns that preserve tissue self-identity, regulate immunity, and control growth. We hypothesize that this coordination undergoes a structured collapse during cancer progression and that there is a shared point of no return across multiple cancer types that may help predict patient outcomes. We calculated gene-gene coordination across 8,314 tumors spanning 32 cancer types by measuring how similarly each pair of 3,000 genes behaves across the patient population. We find that 71% of all coordination is captured by a single pattern, with three secondary patterns that predict survival in 11 of 30 cancer types including patterns involving metabolic dedifferentiation (cytochrome P450 metabolism vs. epithelial differentiation), immune polarization (adaptive vs. innate immunity), and tissue selfidentity. Specifically, this method stratifies prognosis in glioblastoma (C-index 0.845, p = 1.8 x 10-6), prostate adenocarcinoma (0.800), clear cell renal carcinoma (0.649), and eight additional cancers (p = 6.9 x 10-32). Beyond this threshold, tumors of all types share uniformly poor survival, defining a universal molecular point of no return and a potential window for early intervention before coordination collapse becomes irreversible.
Matching journals
The top 7 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Limited inhibition of multiple nodes in a driver network blocks metastasis 94%
- DUX4 is a common driver of immune evasion and immunotherapy failure in metastatic cancers 94%
- MGPfactXMBD: A Model-Based Factorization Method for scRNA Data Unveils Bifurcating Transcriptional Modules Underlying Cell Fate Determination 93%
Similar papers in this journal
Similar papers in this journal
- A tissue-aware machine learning framework enhances the mechanistic understanding and genetic diagnosis of Mendelian and rare diseases 93%
- Explainable Machine Learning Identifies Dosage Compensation Factors in Aneuploid Human Cancer Cells 93%
- KDML: a machine-learning framework for inference of multi-scale gene functions from genetic perturbation screens 93%
Similar papers in this journal
- Cancer Hallmarks Define a Continuum of Plastic Cell States between Small Cell Lung Cancer Archetypes 95%
- Differential Allele-Specific Expression Uncovers Breast Cancer Genes Dysregulated By Cis Noncoding Mutations 94%
- Uncovering the spatial landscape of molecular interactions within the tumor microenvironment through latent spaces 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.