Machine Learning Modeling for Ultrasonication-Mediated Fermentation of Penicillium brevicompactum to Enhance the Release of Mycophenolic Acid.

Described here is the modeling used to improve the mycophenolic acid (MPA) titer from Penicillium brevicompactum using central composite design and a comparatively newer, data-centric approach method k-nearest-neighbor algorithm. The two models for enhancing MPA production using P. brevicompactum were compared with respect to ultrasonic stimulation. During the ultrasonic treatment, we studied different independent factors such as ultrasound power, irradiation duration, treatment frequency and duty cycle to determine their ability to enhance the MPA titer value. The optimized factors such as a treatment time of 10 min (50% duty cycles) with a 12-h interlude at fixed ultrasonic power and frequency (200 W, 40 kHz) were used for ultrasonic treatment of a mycelial culture from the 2nd to 10th day of fermentation. Thus the production of MPA was improved 1.64-fold under the optimized sonication conditions compared with the non-sonicated batch fermentation (non-optimized conditions).

Machine learning modelling for the ultrasonication-mediated disruption of recombinant E. coli for the efficient release of nitrilase.

The ultrasonication-mediated cell disruption of recombinant E. coli was modeled using three machine learning techniques namely Multiple linear regression (MLR), Multi-layer perceptron (MLP) and Sequential minimal optimization (SMO). The four attributes were cellmass concentration (g/L), acoustic power (A), duty cycle (%) and treatment time of sonication (min). For the three responses (nitrilase, total protein release and cell disruption) MLP model was found to be at par with RSM model in terms of generalization as well as prediction capability. Nitrilase release was significantly influenced by the cellmass concentration so was in case of total protein release. Fraction of cells disrupted was heavily influenced by acoustic power and sonication time. Almost 32 U/mL nitrilase could be released for 300 g/L cellmass concentration when sonicated at 225 W for 1 min with 20% duty cycle.

ChemSuite: A package for chemoinformatics calculations and machine learning.

Prediction of biological and toxicological properties of small molecules using in silico approaches has become a wide practice in pharmaceutical research to lessen the cost and enhance productivity. The development of a tool "ChemSuite," a stand-alone application for chemoinformatics calculations and machine-learning model development, is reported. Availability of multi-functional features makes it widely acceptable in various fields. Force field such as UFF is incorporated in tool for optimization of molecules. Packages like RDKit, PyDPI and PaDEL help to calculate 1D, 2D and 3D descriptors and more than 10 types of fingerprints. MinMax Scaler and Z-Score algorithms are available to normalize descriptor values. Varied descriptor selection and machine-learning algorithms are available for model development. It allows the user to add their own algorithm or extend the software for various scientific purposes. It is free, open source and has user-friendly graphical interface, and it can work on all major platforms.

Predicting Inhibitors for Multidrug Resistance Associated Protein-2 Transporter by Machine Learning Approach.

BACKGROUND: The efflux transporter multidrug resistance associated protein-2 belongs to ATP-binding cassette superfamily which plays an important role in multidrug resistance and drugdrug interactions. Efflux transporters are considered to be important targets for increasing the efficacy of drugs and importance of computational study of efflux transporters for predicting substrates, non-substrates, inhibitors and non-inhibitors is well documented. Previous work on predictive models for inhibitors of multidrug resistance associated Protein-2 efflux transporter showed that machine learning methods produced good results. OBJECTIVE: The aim of the present work was to develop a machine learning predictive model to classify inhibitors and non-inhibitors of multidrug resistance associated protein-2 transporter using a well refined dataset. METHOD: In this study, the various algorithms of machine learning were used to develop the predictive models i.e. support vector machine, random forest and k-nearest neighbor. The methods like variance threshold, SelectKBest, random forest, and recursive feature elimination were used to select the features generated by PyDPI. A total of 239 molecules consisting of 124 inhibitors and 115 non-inhibitors were used for model development. RESULTS: The best multidrug resistance associated protein-2 inhibitor model showed prediction accuracies of 0.76, 0.72 and 0.79 for training, 5-fold cross-validation and external sets, respectively. CONCLUSION: It was observed that support vector machine model built on features selected using recursive feature elimination method shows the best performance. The developed model can be used in the early stages of drug discovery for identifying the inhibitors of multidrug resistance associated protein-2 efflux transporter.

Ultrasonic disruption of Pseudomonas putida for the release of arginine deiminase: Kinetics and predictive models.

The responses of the ultrasound-mediated disruption of Pseudomonas putida KT2440 were modelled as the function of biomass concentration in the cell suspension; the treatment time of sonication; the duty cycle and the acoustic power of the sonicator. For the experimental data, the response surface (RSM), the artificial neural network (ANN) and the support vector machine (SVM) models were compared for their ability to predict the performance parameters. The satisfactory prediction of the unseen data of the responses implied the proficient generalization capabilities of ANN. The extent of the cell disruption was mainly dependent on the acoustic power and the biomass concentration. The cellmass concentration in the slurry most strongly influenced the ADI and total protein release. Nearly 28U/mL ADI was released when a biomass concentration of 300g/L was sonicated for 6min with an acoustic power of 187.5W at 40% duty cycle. Cell disruption obeyed first-order kinetics.

The role of multidrug resistance protein (MRP-1) as an active efflux transporter on blood-brain barrier (BBB) permeability.

Drugs acting on central nervous system (CNS) may take longer duration to reach the market as these compounds have a higher attrition rate in clinical trials due to the complexity of the brain, side effects, and poor blood-brain barrier (BBB) permeability compared to non-CNS-acting compounds. The roles of active efflux transporters with BBB are still unclear. The aim of the present work was to develop a predictive model for BBB permeability that includes the MRP-1 transporter, which is considered as an active efflux transporter. A support vector machine model was developed for the classification of MRP-1 substrates and non-substrates, which was validated with an external data set and Y-randomization method. An artificial neural network model has been developed to evaluate the role of MRP-1 on BBB permeation. A total of nine descriptors were selected, which included molecular weight, topological polar surface area, ClogP, number of hydrogen bond donors, number of hydrogen bond acceptors, number of rotatable bonds, P-gp, BCRP, and MRP-1 substrate probabilities for model development. We identified 5 molecules that fulfilled all criteria required for passive permeation of BBB, but they all have a low logBB value, which suggested that the molecules were effluxed by the MRP-1 transporter.

Integrated Computational Solution for Predicting Skin Sensitization Potential of Molecules.

INTRODUCTION: Skin sensitization forms a major toxicological endpoint for dermatology and cosmetic products. Recent ban on animal testing for cosmetics demands for alternative methods. We developed an integrated computational solution (SkinSense) that offers a robust solution and addresses the limitations of existing computational tools i.e. high false positive rate and/or limited coverage. RESULTS: The key components of our solution include: QSAR models selected from a combinatorial set, similarity information and literature-derived sub-structure patterns of known skin protein reactive groups. Its prediction performance on a challenge set of molecules showed accuracy = 75.32%, CCR = 74.36%, sensitivity = 70.00% and specificity = 78.72%, which is better than several existing tools including VEGA (accuracy = 45.00% and CCR = 54.17% with 'High' reliability scoring), DEREK (accuracy = 72.73% and CCR = 71.44%) and TOPKAT (accuracy = 60.00% and CCR = 61.67%). Although, TIMES-SS showed higher predictive power (accuracy = 90.00% and CCR = 92.86%), the coverage was very low (only 10 out of 77 molecules were predicted reliably). CONCLUSIONS: Owing to improved prediction performance and coverage, our solution can serve as a useful expert system towards Integrated Approaches to Testing and Assessment for skin sensitization. It would be invaluable to cosmetic/ dermatology industry for pre-screening their molecules, and reducing time, cost and animal testing.