Background Recent multi‑omics studies show that integrating genomics, transcriptomics, proteomics, and metabolomics improves cancer detection and therapy selection (AI-driven multi-omics integration in precision oncology)[https://pubmed.ncbi.nlm.nih.gov/41266662/]. Yet, RNA and protein layers often disagree, limiting model interpretability (Strategies for Comprehensive Multi-Omics Integrative Data Analysis)[https://pmc.ncbi.nlm.nih.gov/articles/PMC11592251/]. In gastric cancer, transcript‑protein‑metabolite stratification revealed HER2‑linked metabolic subtypes (Integration of transcriptomics, proteomics, and metabolomics data to reveal HER2-associated metabolic heterogeneity)[https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2022.951137/full]. These observations suggest that post‑transcriptional buffering mechanisms shape the observable proteome.

Hypothesis We hypothesize that metabolite‑guided chaperone networks actively buffer transcriptomic fluctuations, producing a stable proteomic layer that decouples mRNA abundance from protein levels in a context‑dependent manner. This buffering is mediated by specific oncometabolites (e.g., 2‑hydroxyglutarate, succinate) that allosterically regulate HSP90 and HSP70 activity, thereby stabilizing a subset of oncogenic proteins despite divergent transcript signals.

Mechanistic Rationale

1. Metabolite‑Chaperone Interaction – Structural studies indicate that certain TCA‑cycle intermediates bind the nucleotide‑binding pocket of HSP90, altering its ATPase cycle and client affinity. Elevated succinate in hypoxic tumors, for example, stabilizes HIF‑1α via HSP90 inhibition, independent of HIF1A mRNA changes.
2. Feedback Loop – Chaperone activity influences metabolic enzyme stability, creating a reciprocal loop where chaperone‑client complexes modulate local metabolite concentrations.
3. Network Specificity – Not all proteins are equally buffered; clients with intrinsically disordered regions or specific motifs show higher chaperone dependence, generating a subset of discordant RNA‑protein pairs.
4. Temporal Dimension – Because metabolite pools shift faster than transcriptional programs, chaperone buffering provides a rapid post‑transcriptional buffer that can be captured in longitudinal multi‑omics snapshots.

Testable Predictions

• Perturbing intracellular succinate or 2‑HG levels will change the phosphorylation state and client load of HSP90/HSP70 without altering global HSP90 expression.
• Proteins identified as high‑confidence chaperone clients in public datasets will show lower correlation between mRNA and protein levels across tumor samples, especially in metabolite‑rich subtypes.
• Inhibiting HSP90 ATPase activity will increase the variance of protein levels relative to mRNA, worsening the RNA‑protein discordance metric used in multi‑omics models.
• Metabolite supplementation will rescue protein levels of specific oncogenic clients after transcriptional knockdown, demonstrating buffering capacity.

Experimental Design

1. Cell‑Line Models – Use isogenic gastric cancer lines with defined HER2 status. Treat with cell‑permeable succinate dimethyl ester or 2‑HG, and measure intracellular metabolite concentrations via LC‑MS.
2. Chaperone Activity Assay – Perform HSP90 ATPase assays and co‑immunoprecipitation to quantify client load under metabolite perturbations.
3. Multi‑Omics Profiling – Collect matched RNA‑seq, quantitative proteomics (TMT), and untargeted metabolomics at 0, 6, 12, 24 h post‑treatment.
4. Discordance Analysis – Compute RNA‑protein Pearson correlation per gene; compare distribution shifts between control and metabolite‑treated conditions.
5. Rescue Experiments – siRNA knockdown of a target oncogene (e.g., EGFR) followed by metabolite supplementation; assess protein recovery by Western blot.
6. In‑Silico Validation – Integrate publicly available TCGA gastric cancer multi‑omics datasets to test whether metabolite‑subtype labels predict RNA‑protein discordance scores.

Potential Pitfalls and Alternatives

• Off‑target effects of metabolite analogues could chaperone-independent signaling; control with structurally inactive analogues.
• Compensatory upregulation of other chaperones may mask HSP90‑specific effects; employ isoform‑specific inhibitors or CRISPR knock‑down.
• If metabolite changes do not alter chaperone client load, the hypothesis would be falsified, prompting investigation of alternative buffering mechanisms such as RNA‑binding proteins or microRNA‑mediated translational control.

Implications Confirming metabolite‑guided chaperone buffering would provide a mechanistic explanation for a major source of noise in multi‑omics cancer models. It would also suggest that incorporating metabolite‑chaperone interaction scores into AI‑driven integration frameworks could improve predictive accuracy and biological interpretability, moving us closer to reliable N‑of‑1 clinical decision tools.