SINGH, A. | bioRxiv Science

bioRxiv2025

Unveiling the structural insights of PFAS-β-lactoglobulin binding mechanism mediating neuronal toxicity in neonates.

The strong, polar-covalent nature of C-F bonds contributes to the forever nature of per- and polyfluoroalkyl (PFAS) substances. PFAS are toxic to humans. Here, we have examined the ability of the small, globular, milk protein {beta}-lactoglobulin to bind PFAS. The protein transports hydrophobic and amphiphilic compounds, including retinol and fatty acids, for vision and brain development; therefore, underscoring its interactions with PFAS is significant. The crystal structures of {beta}-lactoglobulin complexed with PFOA (Perfluorooctanoic acid) at 2.0 ([A]), PFOS (Perfluorooctanesulfonic acid) at 2.5 ([A]), and PFDA (Perfluorodecanoic acid) at 2.0 ([A]) reveal high affinity of the compounds for the central calyx of {beta}-lactoglobulin, which is the canonical retinol and fatty acid binding site. Analyses of the data indicate significant hydrophobic interactions stabilizing the binding of the PFAS hydrophobic "tails" within the calyx and interactions between Lys60 and Lys69 and PFAS polar head groups. Comparative structural analysis revealed the presence of an open conformation of the EF loop containing the Glu89 latch residue in the complexed structures vis-a-vis the apo-form. Molecular dynamics (MD) simulations revealed high stability of the PFAS binding and attainment of energy minima in all complexes. The average binding energy of PFDA in {beta}-lactoglobulin calyx was -25 kcal/mol, which was higher than PFOS (-21 kcal/mol) and PFOA (-23 kcal/mol) due to increased van der Waals interactions of the longer hydrophobic chain of PFDA with {beta}-lactoglobulin. This work advances a mechanism by which {beta}-lactoglobulin can recruit PFAS and act as a transporter for the "forever" chemical, potentially mediating its neurotoxicity.

biophysics↗

bioRxiv2021

Amino Acid Composition and Charge Based Prediction of Antisepsis Peptides by Random Forest Machine Learning Algorithm

Sepsis is a severe infectious disease with high mortality, and it occurs when chemicals released in the bloodstream to fight an infection trigger inflammation throughout the body and it can cause a cascade of changes that damage multiple organ systems, leading them to fail, even resulting in death. In order to reduce the possibility of sepsis or infection antiseptics are used and process is known as antisepsis. Antiseptic peptides (ASPs) show properties similar to antigram-negative peptides, antigram-positive peptides and many more. Machine learning algorithms are useful in screening and identification of therapeutic peptides and thus provide initial filters or built confidence before using time consuming and laborious experimental approaches. In this study, various machine learning algorithms like Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbour (KNN) and Logistic Regression (LR) were evaluated for prediction of ASPs. Moreover, the characteristics physicochemical features of ASPs were also explored to use them in machine learning. Both manual and automatic feature selection methodology was employed to achieve best performance of machine learning algorithms. A 5-fold cross validation and independent data set validation proved RF as the best model for prediction of ASPs. Our RF model showed an accuracy of 97%, Matthews Correlation Coefficient (MCC) of 0.93, which are indication of a robust and good model. To our knowledge this is the first attempt to build a machine learning classifier for prediction of ASPs.

bioinformatics↗