UNIKIN Shines in Athens (Greece): Prof. Selain Kasereka Presents Two Scientific Papers at the MobiSPC 2026 International Conference

Published in the journal Procedia Computer Science (Elsevier / ScienceDirect), these two works illustrate the vitality of research conducted by researchers of Congolese origin on an international scale and demonstrate the scientific potential that the University of Kinshasa continues to project beyond its borders.

The first paper, titled PG–NODETB: Physics-Guided Neural Ordinary Differential Equations for Tuberculosis Transmission Dynamics, introduces an innovative framework combining physics-guided ordinary differential equations and artificial neural networks—a PG-NODE approach—to model tuberculosis (TB) transmission dynamics.

Facing the limitations of classical compartmental models, which are rigid due to their fixed-parameter structure, the authors propose an enriched SLIR model capable of learning time-varying transmission functions while preserving biological conservation laws. A rigorous mathematical analysis of the basic reproduction number (R_0), equilibrium stability, and sensitivity indices was conducted. Three simulation scenarios were examined: (i) adaptive tracking of non-stationary transmission rates; (ii) a 27% reduction in root mean square error compared to classical SLIR models; and (iii) multi-lever health policy optimization bringing R_0 down to 1.49 under combined interventions.

This work opens particularly promising prospects for connected health environments, where mobile platforms generate continuous surveillance data—a growing reality in resource-limited African countries.

The second paper, titled BioMpox: A Biologically Informed Compartment Model for Mpox Transmission Dynamics, addresses the modeling of Mpox transmission—a re-emerging zoonosis whose epidemic dynamics are often poorly captured by traditional compartmental models.

The authors propose a refined SEPIRD model incorporating a pre-symptomatic progression stage and a general biological transition function ϕ(P,θ), allowing for a more realistic representation of disease progression processes. Analytically and numerically compared to a classical SEIRD model, BioMpox produces a lower and delayed infectious peak—1.15% at hour 242 compared to 12.4% at hour 68 for the classical model—representing an almost ten-fold reduction in peak infectious prevalence without altering the final attack size.

These findings have direct implications for hospital capacity planning and the design of intervention strategies in epidemic contexts, particularly in Central Africa, where Mpox remains a major public health threat.