Evaluation of a neural networks QSAR method based on ligand representation using substituent descriptors. Application to HIV-1 protease inhibitors

Predicting deep well pump performance with machine learning methods during hydraulic head changes

In this study, machine learning techniques were employed to estimate and predict the system efficiency of a pumping plant at various hydraulic head levels. The measured parameters, including flow rate, outlet pressure, drawdown, and power, were used for estimating the system efficiency. Two approaches, Approach-I and Approach-II, were utilized. Approach-I incorporated additional parameters such as hydraulic head, drawdown, flow, power, and outlet pressure, while Approach-II focused solely on hydraulic head, outlet pressure, and power. Seven machine learning algorithms were employed to model and predict the efficiency of the pumping plant. The decrease in the hydraulic head by 125 cm resulted in a reduction in the pump system efficiency by 6.45 %, 8.94 %, and 13.8 % at flow rates of 40, 50, and 60 m3 h-1, respectively. Among the algorithms used in Approach-I, the artificial neural network, support vector machine regression, and lasso regression exhibited the highest performance, with R2 values of 0.995, 0.987, and 0.985, respectively. The corresponding RMSE values for these algorithms were 0.13 %, 0.23 %, and 0.22 %, while the MAE values were 0.11 %, 0.2 %, and 0.32 %, and the MAPE values were 0.22 %, 0.5 %, and 0.46.% In Approach-II, the artificial neural network model once again demonstrated the best performance with an R2 value of 0.996, followed by the support vector machine regression (R2 = 0.988) and the decision tree regression (R2 = 0.981). Overall, the artificial neural network model proved to be the most effective in both approaches. These findings highlight the potential of machine learning techniques in predicting the efficiency of pumping plant systems.

Learning the Relationship between the Primary Structure of HIV Envelope Glycoproteins and Neutralization Activity of Particular Antibodies by Using Artificial Neural Networks

The dependency between the primary structure of HIV envelope glycoproteins (ENV) and the neutralization data for given antibodies is very complicated and depends on a large number of factors, such as the binding affinity of a given antibody for a given ENV protein, and the intrinsic infection kinetics of the viral strain. This paper presents a first approach to learning these dependencies using an artificial feedforward neural network which is trained to learn from experimental data. The results presented here demonstrate that the trained neural network is able to generalize on new viral strains and to predict reliable values of neutralizing activities of given antibodies against HIV-1.