Back

Using AI-MIDD for Mechanistic Erythropoietin Modeling: A Digital-Twin Framework for Optimizing ESA Therapies

Goryanin, I.; Goryanin, I. V.

2025-11-13 systems biology
10.1101/2025.11.12.688044 bioRxiv
Show abstract

Erythropoiesis-stimulating agents (ESAs) remain a mainstay for anemia therapy, yet variability in response and emerging resistance mechanisms limit their effectiveness. We developed a hybrid AI-MIDD (Model-Informed Drug Development) and Quantitative Systems Pharmacology (QSP) platform and applied it for integrating mechanistic signaling (EPO-EpoR-JAK2/STAT5-SOCS3/CIS) with metabolic (mTOR), iron-homeostatic (hepcidin), and HIF-mediated endogenous EPO feedback. The model was implemented in SBML (epo_qsp_combined_all.xml) and simulated over 12 weeks under various ESA, SUMO-blocker, miR-486 exosome, mTOR, and HIF-PHI perturbations. AI-assisted parameter scanning revealed distinct dose-sparing regimes: SUMO inhibition improved receptor recycling and reduced ESA requirement by [~]30%; exosomal miR-486 reduced SOCS3/CIS burden, restoring STAT5 sensitivity; HIF-PHI enhanced baseline EPO synthesis, while mTOR modulation stabilized reticulocyte oscillations. Multi-objective optimization identified triplet combinations achieving Hb targets with minimized pSTAT5 burden. The AI-MIDD-QSP integration provides a digital-twin for patient-specific ESA optimization and enables rational design of combination regimens and patentable therapeutic concepts. This framework generalizes to other hematopoietic and cytokine-signaling systems, advancing mechanism-based drug development. Manuscript HighlightsO_LIA hybrid AI-MIDD and QSP "digital-twin" of erythropoiesis was developed, integrating core EPO signaling with SUMO-recycling, exosomal miR-486, and mTOR metabolic pathways. C_LIO_LIThe model identifies and quantifies SUMO-pathway inhibition as a novel, dose-sparing mechanism, showing a [~]30% reduction in ESA requirement by increasing EpoR membrane recycling. C_LIO_LIThe model demonstrates how exosomal miR-486 delivery can restore signaling sensitivity in resistant states by reducing the SOCS3/CIS negative feedback burden by [~]40%. C_LIO_LIAI-driven multi-objective optimization identified a novel triplet combination (ESA + SUMO inhibition + miR-486) as the most effective regimen, achieving target hemoglobin with a 45% reduction in cumulative ESA and minimized pSTAT5 signaling burden. C_LI

Matching journals

The top 12 journals account for 50% of the predicted probability mass.

50% of probability mass above

"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.