ARTIFICIAL INTELLIGENCE AND CHEMINFORMATICS STRATEGIES FOR OPTIMIZING R&D INVESTMENT DECISIONS IN THE RADIOPHARMACEUTICAL INDUSTRY THROUGH ECONOMIC POTENTIAL AND AN INTEGRATED GOVERNANCE FRAMEWORK

Hoang, Sy Minh Tuan

doi:10.22038/aojnmb.2026.98071.1742

ARTIFICIAL INTELLIGENCE AND CHEMINFORMATICS STRATEGIES FOR OPTIMIZING R&D INVESTMENT DECISIONS IN THE RADIOPHARMACEUTICAL INDUSTRY THROUGH ECONOMIC POTENTIAL AND AN INTEGRATED GOVERNANCE FRAMEWORK

Articles in Press

Document Type : Original Article

Author

Sy Minh Tuan Hoang

Thu Dau Mot University, Ho Chi Minh City, Vietnam

10.22038/aojnmb.2026.98071.1742

Abstract

The advanced pharmaceutical and biomedical industry faces a profound financial paradox: as analytical technology capability grows exponentially, R&D investment efficiency continues to decline under Eroom's Law, a decline that is especially pronounced in the radiopharmaceutical development subsector involving the heavy-metal isotopes Lutetium–177 (Lu–177) and Yttrium–90 (Y–90). This paper presents a Decision Support System (DSS) framework that integrates cheminformatics and artificial intelligence (AI), shifting the research paradigm from wet-chemistry “trial-and-error” toward an in silico simulation environment. Using techno-economic analysis (TEA) and total cost of ownership (TCO) measurement that accounts for GMP-compliant operational risk, the study models the discovery-stage investment. The core academic contribution lies in coupling graph neural networks (GNN) with Evidential Deep Learning and quantum data from density functional theory (DFT), using norm-conserving pseudopotentials under the GTH scheme with 4f electrons relegated to the atomic core (4f–in–core) (1,2). The study quantifies the capacity of the AI system to cut the discovery budget by approximately 67% and compress the preclinical candidate design cycle by 60–70%. This system not only predicts thermodynamic stability constants but also establishes a stringent filter for kinetic inertness, satisfying the full preclinical requirements for an Investigational New Drug (IND) application to the U.S. Food and Drug Administration (FDA).

Keywords

Radiopharmaceuticals

Cheminformatics

Evidential Deep Learning

4f–in–core pseudopotentials

Techno-Economic Analysis

Subjects