Spatial Multi-Omics Discordance Predicts Early Therapy Resistance in Solid Tumors

Hypothesis

Integrated spatial proteogenomic and metabolomic profiles reveal layer‑specific discordance that precedes clinical resistance to targeted therapy, and measuring this discordance predicts non‑response earlier than bulk multi‑omics signatures.

Mechanistic Rationale

Single‑cell proteomics shows that protein abundance often diverges from transcript levels due to post‑translational modifications, while metabolomics reflects pathway flux. In tumor microenvironments, stromal cues induce rapid phosphorylation of signaling proteins without corresponding mRNA changes, altering metabolic enzyme activity and creating a mismatch between the proteomic and metabolomic layers. Bulk averaging masks these transient shifts, but spatial preservation captures them. Recent multi‑omics studies demonstrate that organ‑specific plasma proteome links to imaging phenotypes invisible in genomics alone (Organ‑specific plasma proteome links), and that integrated biological age predictors outperform single‑layer clocks by capturing system‑level damage (Integrated biological age predictors). Extending this logic, early therapeutic pressure should produce a detectable proteome‑metabolome discordance within malignant niches before genomic alterations or bulk expression shifts appear.

Testable Predictions

1. It's expected that in patients receiving EGFR‑inhibitor therapy for non‑small cell lung cancer, a rising proteome‑metabolome discordance score in circulating tumor‑derived exosomes will precede radiographic progression by at least 4‑6 weeks.
2. Patients whose baseline discordance score falls in the top quartile will exhibit a hazard ratio >2.0 for progression‑free survival compared with those in the bottom quartile, independent of TMB or PD‑L1 status.
3. Experimental induction of kinase inhibition in organoid cultures will generate a measurable increase in phospho‑proteome signal without matching changes in metabolite pools, which can be reversed by combined phosphatase inhibition.

Experimental Design

• Cohort: 120 treatment‑naïve NSCLC patients scheduled for first‑line osimertinib; collect peripheral blood at baseline, weekly for 8 weeks, and at clinical progression.
• Isolation: Exosome‑enriched fractions; perform spatial proteomics using multiplexed ion beam imaging (MIBI) on exosome‑derived protein arrays and untargeted metabolomics via LC‑MS.
• Discordance score: Compute the absolute deviation between normalized protein abundance pathways (e.g., MAPK, PI3K) and corresponding metabolite ratios (e.g., ATP/ADP, NADH/NAD+), then aggregate across pathways using a weighted Euclidean distance.
• Statistical analysis: Time‑dependent ROC curves to assess prediction of progression; Cox proportional hazards models adjusting for age, smoking status, and baseline tumor burden.
• Validation: Parallel patient‑derived organoid treated with osimertinib; sample at 0, 6, 12, 24 h; perform phospho‑proteomics and targeted metabolomics; test reversal with broad‑spectrum phosphatase inhibitor.
• Controls: Spike‑in standards correct for exosome heterogeneity and paired plasma metabolomics controls for circulating metabolite flux.

If the discordance score fails to anticipate progression in the prespecified window, the hypothesis is falsified. Conversely, a significant lead time would support the notion that early multi‑omics layer mismatch captures functional drug resistance mechanisms invisible to genomics or bulk omics alone, guiding adaptive combination therapies before clinical relapse.