Artificial Neural Network Modeling to Predict Bacterial Attachment on Composite Biopolymeric Scaffold

Abstract

Biomaterial-associated infection is a significant cause of concern in health care and clinical field. It is widely acknowledged that material with intended use in a medical application must not support colonization and attachment of bacteria on its surfaces. This has encouraged the search for the newer polymeric biomaterial with anti-adhesive and anti-adherence properties to resist bacterial attachment. Firsthand information about the combined effect of various material properties that affects bacterial attachment to biomaterial surface during the development of anti-adhesive material is always useful. In this work, computer-aided techniques, i.e., the neural network was applied to build a model using biomaterial properties such as surface roughness, swelling ratio, and pore size as input and number of surface adhered bacterial cells as output response. Various blending combinations of silk fibroin and xanthan were used to generate biocomposites with varying surface roughnesses. The surface roughness of the composite scaffold was determined through Atomic Force Microscopy (AFM), and surface parameter was evaluated using inbuilt software provided with AFM, and the number of bacteria attached to biocomposite surface was estimated through ImageJ. The coefficient of determination (R2) and Mean Absolute Percentage Error (MAPE) of the neural model were determined as 0.990 and 0.26%, respectively. Thus, confirming that the developed model has excellent prediction accuracy. This model, when used for prediction with the entirely unseen dataset, gave R2 as 0.982, pointing towards the robustness of model in predicting the similar type of system within the limit of the trained data set. © 2018, Springer Nature Singapore Pte Ltd.