Efficient prediction of drug–drug interaction using deep learning models

A drug–drug interaction or drug synergy is extensively utilised for cancer treatment. However, prediction of drug–drug interaction is defined as an ill‐posed problem, because manual testing is only implementable on small group of drugs. Predicting the drug–drug interaction score has been a popular research topic recently. Recently many machine learning models have proposed in the literature to predict the drug–drug interaction score efficiently. However, these models suffer from the over‐fitting issue. Therefore, these models are not so‐effective for predicting the drug–drug interaction score. In this work, an integrated convolutional mixture density recurrent neural network is proposed and implemented. The proposed model integrates convolutional neural networks, recurrent neural networks and mixture density networks. Extensive comparative analysis reveals that the proposed model significantly outperforms the competitive models.Inspec keywords: cancer, learning (artificial intelligence), drugs, recurrent neural nets, convolutional neural nets, drug delivery systemsOther keywords: drug synergy, drug–drug interaction score, drug–drug interaction prediction, deep learning, cancer treatment, machine learning, convolutional mixture density recurrent neural network     

Effect of hook characteristics on the fracture behaviour of dissimilar friction stir welded aluminium alloy and mild steel sheets

ABSTRACT

Under tensile shear loading, fracture modes of dissimilar lap welds produced by friction stir scribe technology were studied. Three fracture modes were observed. For zone A fracture, the initial crack was restrained, and the joint ultimately fractured in the base mild steel. For zone B fracture, the initial crack progressed through the aluminium sheet just above the Al/steel interface. For zone C fracture, the initial crack proceeded along the steel hook to the aluminium sheet surface. Fracture mode and joint strength were greatly influenced by steel hook size, and the steel hook size was affected by welding parameters and tool scribe height. In this study, the experimental joint strength achieved the calculated joint load limit.     

Thermal Decomposition Of Carbonyl Sulfide At Temperatures Encountered In The Front End Of Modified Claus Plants

Understanding the kinetics of the formation and consumption of COS and CS₂ in the front end of the modified Claus process will be a significant step towards reducing the environmental impact of these plants. Specifically, homogeneous intrinsic rate expressions are needed for engineering design and simulation, which will lead to new, optimized ways of operating these plants. Hence, a high-temperature kinetic study of the COS decomposition reaction was carried out. Experiments were performed with inlet COS compositions in the range of 0.20-2.33 mol%, with pressures at 101-150 kPa and temperatures at 800-1100°C; these conditions cover the conditions typically encountered in the front end of the modified Claus process. The experimental results showed that COS conversion is dependent on the inlet concentration of COS, which contrasts with previously reported higher temperature studies. Finally, the COS decomposition kinetics were modeled as the sum of two reactions, which provided a satisfactory representation of experimental data.     

Serological distinction of integral plasma membrane proteins as a class of mycobacterial antigens and their relevance for human T cell activation

This study pertains to classification and antigenic analysis of mycobacterial plasma membrane proteins in relation to human T cell proliferative responses, using a 'fast grower' Mycobacterium fortuitum as model. Membrane vesicles, prepared by sonication and differential centrifugation, were subjected to biphasic Triton X-114 extraction for isolation of integral peripheral (aqueous phase) proteins. Neither protein pool showed any appreciable overlap serologically. SDS-PAGE showed five prominent bands in peripheral and three in the integral protein pool, whereas immunoblotting with rabbit antisera identified only two major antigens (60 and 67 kD) in the former and five (24, 34, 42, 51 and 54 kD) in the latter. ELISA with a panel of anti-mycobacterial MoAbs revealed that nine out of 12 previously known antigens were present in the peripheral protein pool. Only two of them (33 and 40 kD) were additionally detected amongst integral proteins. The membrane-associated immunosuppressive moiety lipoarabinomannan was semiquantitatively located in aqueous phase. In bulk T cell proliferation assays, seven out of 10 subjects belonging to a 'responder' background (BT-BB leprosy patients and healthy contacts) showed high responses for Myco. fortuitum antigens. Proliferative response with integral proteins was comparable to that with whole membrane, but it was significantly higher (P < 0.0005) than the response with peripheral proteins. The distinction and relevance of integral membrane proteins as a class of mycobacterial antigens make them worthy of consideration in a subunit vaccine design.     

X-ray photoelectron spectroscopic studies of the effect of laser irradiation on the oxidation behaviour of modified 9Cr-1Mo ferritic steel

DMSRDE Communication No. 2118.     

Economic comparison of current electricity generating technologies and advanced nuclear options

This analysis compares the cost of various electric grid scenarios at the national level over a one-year period. Scenarios include high renewable, zero nuclear, zero carbon, and deployment of advanced nuclear. Additionally, several carbon tax scenarios are explored in the model to further assess the cost generation if the current nuclear fleet retired. The cheapest scenarios were those that involved less reliance on conventional nuclear reactors and more molten salt reactors, renewables, or natural gas.     

Measurement of asphaltene agglomeration from cold lake bitumen diluted with n-alkanes

A laser particle analyzer was used to study the formation of asphaltene agglomerates from Cold Lake bitumen due to the introduction of paraffinic diluents. The growth of the particles over time was investigated, and it was found that asphaltene particle growth is essentially an instantaneous process; that is, the final agglomerate size is reached within 5 seconds. The type of diluent used to precipitate the asphaltenes was determined to have a major effect on the size distribution of asphaltene agglomerates. The average particle size of asphaltene agglomerates ranged from 169 ptn for n-hexadecane to 299m for n-pentane. A relationship is shown to exist between the average particle size of asphaltene agglomerates and the carbon number of the diluent used.