Ahmad Alzghoul

Computational predictions of glass-forming ability and crystallization tendency of drug molecules.

Amorphization is an attractive formulation technique for drugs suffering from poor aqueous solubility as a result of their high lattice energy. Computational models that can predict the material properties associated with amorphization, such as glass-forming ability (GFA) and crystallization behavior in the dry state, would be a time-saving, cost-effective, and material-sparing approach compared to traditional experimental procedures. This article presents predictive models of these properties developed using support vector machine (SVM) algorithm. The GFA and crystallization tendency were investigated by melt-quenching 131 drug molecules in situ using differential scanning calorimetry. The SVM algorithm was used to develop computational models based on calculated molecular descriptors. The analyses confirmed the previously suggested cutoff molecular weight (MW) of 300 for glass-formers, and also clarified the extent to which MW can be used to predict the GFA of compounds with MW < 300. The topological equivalent of Grav3_3D, which is related to molecular size and shape, was a better descriptor than MW for GFA; it was able to accurately predict 86% of the data set regardless of MW. The potential for crystallization was predicted using molecular descriptors reflecting Hückel pi atomic charges and the number of hydrogen bond acceptors. The models developed could be used in the early drug development stage to indicate whether amorphization would be a suitable formulation strategy for improving the dissolution and/or apparent solubility of poorly soluble compounds.

Data stream forecasting for system fault prediction

Competition among todays industrial companies is very high. Therefore, system availability plays an important role and is a critical point for most companies. Detecting failures at an early stage or foreseeing them before they occur is crucial for machinery availability. Data analysis is the most common method for machine health condition monitoring. In this paper we propose a fault-detection system based on data stream prediction, data stream mining, and data stream management system (DSMS). Companies that are able to predict and avoid the occurrence of failures have an advantage over their competitors. The literature has shown that data prediction can also reduce the consumption of communication resources in distributed data stream processing. In this paper different data-stream-based linear regression prediction methods have been tested and compared within a newly developed fault detection system. Based on the fault detection system, three DSM algorithms outputs are compared to each other and to real data. The three applied and evaluated data stream mining algorithms were: Grid-based classifier, polygon-based method, and one-class support vector machines (OCSVM). The results showed that the linear regression method generally achieved good performance in predicting short-term data. (The best achieved performance was with a Mean Absolute Error (MAE) around 0.4, representing prediction accuracy of 87.5%). Not surprisingly, results showed that the classification accuracy was reduced when using the predicted data. However, the fault-detection system was able to attain an acceptable performance of around 89% classification accuracy when using predicted data.

Experimental and Computational Prediction of Glass Transition Temperature of Drugs

Glass transition temperature (Tg) is an important inherent property of an amorphous solid material which is usually determined experimentally. In this study, the relation between Tg and melting temperature (Tm) was evaluated using a data set of 71 structurally diverse druglike compounds. Further, in silico models for prediction of Tg were developed based on calculated molecular descriptors and linear (multilinear regression, partial least-squares, principal component regression) and nonlinear (neural network, support vector regression) modeling techniques. The models based on Tm predicted Tg with an RMSE of 19.5 K for the test set. Among the five computational models developed herein the support vector regression gave the best result with RMSE of 18.7 K for the test set using only four chemical descriptors. Hence, two different models that predict Tg of drug-like molecules with high accuracy were developed. If Tm is available, a simple linear regression can be used to predict Tg. However, the results also suggest that support vector regression and calculated molecular descriptors can predict Tg with equal accuracy, already before compound synthesis.

Data mining of solubility parameters for computational prediction of drug–excipient miscibility

Abstract Computational data mining is of interest in the pharmaceutical arena for the analysis of massive amounts of data and to assist in the management and utilization of the data. In this study, a data mining approach was used to predict the miscibility of a drug and several excipients, using Hansen solubility parameters (HSPs) as the data set. The K-means clustering algorithm was applied to predict the miscibility of indomethacin with a set of more than 30 compounds based on their partial solubility parameters [dispersion forces , polar forces and hydrogen bonding ]. The miscibility of the compounds was determined experimentally, using differential scanning calorimetry (DSC), in a separate study. The results of the K-means algorithm and DSC were compared to evaluate the K-means clustering prediction performance using the HSPs three-dimensional parameters, the two-dimensional parameters such as volume-dependent solubility and hydrogen bonding , and selected single (one-dimensional) parameters. Using HSPs, the prediction of miscibility by the K-means algorithm correlated well with the DSC results, with an overall accuracy of 94%. The prediction accuracy was the same (94%) when the two-dimensional parameters or the hydrogen-bonding (one-dimensional) parameter were used. The hydrogen-bonding parameter was thus a determining factor in predicting miscibility in such set of compounds, whereas the dispersive and polar parameters had only a weak correlation. The results show that data mining approach is a valuable tool for predicting drug–excipient miscibility because it is easy to use, is time and cost-effective, and is material sparing.

Physical stability of drugs after storage above and below the glass transition temperature: Relationship to glass-forming ability.

Graphical abstract

Data Stream Mining for Increased Functional Product Availability Awareness

Functional Products (FP) and Product Service Systems (PSS) may be seen as integrated systems comprising hardware and support services. For such offerings, availability is key. Little research has been done on integrating Data Stream Management Systems (DSMS) for monitoring (parts of) a FP to improve system availability. This paper introduces an approach for how data stream mining may be applied to monitor hardware being part of a Functional Product. The result shows that DSMS have the potential to significantly support continuous availability awareness of industrial systems, especially important when the supplier is to supply a function with certain availability.

Comparing a knowledge-based and a data-driven method in querying data streams for system fault detection : A hydraulic drive system application

The field of fault detection and diagnosis has been the subject of considerable interest in industry. Fault detection may increase the availability of products, thereby improving their quality. Fau ...

Addressing concept drift to improve system availability by updating one-class data-driven models

Data-driven models have been used to detect system faults, thereby increasing industrial system availability. The ability to search data streams while dealing with concept drift are challenges for data-driven models. The objective of this work is to demonstrate a general method to manage concept drift when using one-class data-driven models. The method has been used to develop an automatically retrained and updated polygon-based model. In this paper, the available industrial data allowed for use of one-class data-driven models, and the polygon-based model was selected because it has previously been successful. Possible scenarios that allow one-class data-driven models to be retrained or updated were identified. Based on the identified scenarios, a method to automatically update a polygon-based model online is proposed. The method has been tested and verified using data collected from a Bosch Rexroth Mellansel AB hydraulic drive system. Data representing relevant faults was inserted into the data set in close collaboration with engineers from the company. The results show that the developed polygon-based model method was able to address the concept drift issue and was able to significantly improve the classification accuracy compared to the static polygon-based model. Thereby, the model could significantly improve industrial system availability when applied in the relevant production process. This paper shows that the developed polygon-based model requires small memory space while its updating procedure is simple and fast. Finally, the identified scenarios may be helpful as input for supporting other one-class data-driven models to cope with concept drift, thus increasing the generalizability of the results.