The integration of deep learning (DL) and photonic crystal fiber (PCF)-based multi-analyte detection biosensors is increasingly used to improve measurement specificity in many crucial applications and to enhance detection sensitivity to the micro-range. For the first time in this letter, we introduce the Physics-Informed Neural Network (PINN) for four different multi-analyte biosensors, enabling detection and analysis with a single model. The model was trained on various datasets and input features (IF) for each biosensor. The proposed PINN model shows the lowest mean square error (MSE) of 0.0022(±0.0002) and the mean absolute error (MAE) of 0.0122(±0.0010) . The performance R2 of the PINN model for four different biosensors is 0.9238(±0.0113) , 0.9788(±0.0031) , 0.9560(±0.0112) , and 0.9883(±0.0025) , respectively. A noticeable trend is that larger datasets lead to better model performance. The proposed PINN model is competent and flexible for designing and optimizing four different applied multi-analyte biosensors and can be used to effectively design and optimize other multi-analyte biosensors.