Chemical acute oral toxicity is an important end point in drug design and environmental risk assessment. However, it is difficult to determine by experiments, and in silico methods are hence developed as an alternative. In this study, a comprehensive data set containing 12, 204 diverse compounds with median lethal dose (LD₅₀) was compiled. These chemicals were classified into four categories, namely categories I, II, III and IV, based on the criterion of the U.S. Environmental Protection Agency (EPA). Then several multiclassification models were developed using five machine learning methods, including support vector machine (SVM), C4.5 decision tree (C4.5), random forest (RF), κ-nearest neighbor (kNN), and naïve Bayes (NB) algorithms, along with MACCS and FP4 fingerprints. One-against-one (OAO) and binary tree (BT) strategies were employed for SVM multiclassification. Performances were measured by two external validation sets containing 1678 and 375 chemicals, separately. The overall accuracy of the MACCS-SVM(OAO) model was 83.0% and 89.9% for external validation sets I and II, respectively, which showed reliable predictive accuracy for each class. In addition, some representative substructures responsible for acute oral toxicity were identified using information gain and substructure frequency analysis methods, which might be very helpful for further study to avoid the toxicity.

Although deep neural networks are highly effective, their high computational and memory costs severely challenge their applications on portable devices. As a consequence, low-bit quantization, which converts a full-precision neural network into a low-bitwidth integer version, has been an active and promising research topic. Existing methods formulate the low-bit quantization of networks as an approximation or optimization problem. Approximation-based methods confront the gradient mismatch problem, while optimization-based methods are only suitable for quantizing weights and could introduce high computational cost in the training stage. In this paper, we propose a novel perspective of interpreting and implementing neural network quantization by formulating low-bit quantization as a differentiable non-linear function (termed quantization function). The proposed quantization function can be learned in a lossless and end-to-end manner and works for any weights and activations of neural networks in a simple and uniform way. Extensive experiments on image classification and object detection tasks show that our quantization networks outperform the state-of-the-art methods. We believe that the proposed method will shed new insights on the interpretation of neural network quantization. Our code is available at https://github.com/aliyun/ alibabacloud-quantization-networks.

No abstract

Dropout has been proven to be an effective algorithm for training robust deep networks because of its ability to prevent overfitting by avoiding the co-adaptation of feature detectors. Current explanations of dropout include bagging, naive Bayes, regularization, and sex in evolution. According to the activation patterns of neurons in the human brain, when faced with different situations, the firing rates of neurons are random and continuous, not binary as current dropout does. Inspired by this phenomenon, we extend the traditional binary dropout to continuous dropout. On the one hand, continuous dropout is considerably closer to the activation characteristics of neurons in the human brain than traditional binary dropout. On the other hand, we demonstrate that continuous dropout has the property of avoiding the co-adaptation of feature detectors, which suggests that we can extract more independent feature detectors for model averaging in the test stage. We introduce the proposed continuous dropout to a feedforward neural network and comprehensively compare it with binary dropout, adaptive dropout, and DropConnect on Modified National Institute of Standards and Technology, Canadian Institute for Advanced Research-10, Street View House Numbers, NORB, and ImageNet large scale visual recognition competition-12. Thorough experiments demonstrate that our method performs better in preventing the co-adaptation of feature detectors and improves test performance.

Current treatments for cancer and the central nervous system diseases are limited, partly due to the difficulties posed by the insolubility, poor distribution of drugs among cells and lack of selectivity of drugs, the inability of drugs to cross cellular barriers and blood brain barrier (BBB). Carbon nanotubes (CNTs) possess many distinct properties including good electronic properties, remarkably penetrating capability on the cell membrane, high drug-loading and pH-dependent therapeutic unloading capacities, thermal properties, large surface area and easy modification with molecules, which render them as a suitable candidate to deliver drugs to cancer and brain. CNTs as a drug delivery could achieve a high efficacy, enhance specificity and diminish side effects. Whereas CNTs have been primarily employed in cancer treatment, a few studies have focused on the treatment and diagnosis of the central nervous system diseases using CNTs. Here, we review the current progress of in vitro and in vivo researches of CNTs-based drug delivery to cancer involving CNTs-based tumor-targeted drug delivery systems (DDS), photodynamic therapy (PDT) and photothermal therapy (PTT). Meanwhile, we also review the current progress of in vitro and in vivo researches of CNTs-based drug delivery to brain.

Carcinogenicity is one of the most concerned properties of chemicals to human health, thus it is important to identify chemical carcinogenicity as early as possible. In this study, 829 diverse compounds with rat carcinogenicity were collected from Carcinogenic Potency Database (CPDB). Using six types of fingerprints to represent the molecules, 30 binary and ternary classification models were generated to predict chemical carcinogenicity by five machine learning methods. The models were evaluated by an external validation set containing 87 chemicals from ISSCAN database. The best binary model was developed by MACCS keys and kNN algorithm with predictive accuracy at 83.91 %, while the best ternary model was also generated by MACCS keys and kNN algorithm with overall accuracy at 80.46 %. Furthermore, the best binary and ternary classification models were used to estimate carcinogenicity of tobacco smoke components containing 2251 compounds. 981 ones were predicted as carcinogens by binary classification model, while 110 compounds were predicted as strong carcinogens and 807 ones as weak carcinogens by ternary classification model. The results indicated that our models would be helpful for prediction of chemical carcinogenicity.

No abstract

No abstract