Precise atom manipulation through deep reinforcement learning.

Atomic-scale manipulation in scanning tunneling microscopy has enabled the creation of quantum states of matter based on artificial structures and extreme miniaturization of computational circuitry based on individual atoms. The ability to autonomously arrange atomic structures with precision will enable the scaling up of nanoscale fabrication and expand the range of artificial structures hosting exotic quantum states. However, the a priori unknown manipulation parameters, the possibility of spontaneous tip apex changes, and the difficulty of modeling tip-atom interactions make it challenging to select manipulation parameters that can achieve atomic precision throughout extended operations. Here we use deep reinforcement learning (DRL) to control the real-world atom manipulation process. Several state-of-the-art reinforcement learning (RL) techniques are used jointly to boost data efficiency. The DRL agent learns to manipulate Ag adatoms on Ag(111) surfaces with optimal precision and is integrated with path planning algorithms to complete an autonomous atomic assembly system. The results demonstrate that state-of-the-art DRL can offer effective solutions to real-world challenges in nanofabrication and powerful approaches to increasingly complex scientific experiments at the atomic scale.

RNA secondary structure prediction with convolutional neural networks.

BACKGROUND: Predicting the secondary, i.e. base-pairing structure of a folded RNA strand is an important problem in synthetic and computational biology. First-principle algorithmic approaches to this task are challenging because existing models of the folding process are inaccurate, and even if a perfect model existed, finding an optimal solution would be in general NP-complete. RESULTS: In this paper, we propose a simple, yet effective data-driven approach. We represent RNA sequences in the form of three-dimensional tensors in which we encode possible relations between all pairs of bases in a given sequence. We then use a convolutional neural network to predict a two-dimensional map which represents the correct pairings between the bases. Our model achieves significant accuracy improvements over existing methods on two standard datasets, RNAStrAlign and ArchiveII, for 10 RNA families, where our experiments show excellent performance of the model across a wide range of sequence lengths. Since our matrix representation and post-processing approaches do not require the structures to be pseudoknot-free, we get similar good performance also for pseudoknotted structures. CONCLUSION: We show how to use an artificial neural network design to predict the structure for a given RNA sequence with high accuracy only by learning from samples whose native structures have been experimentally characterized, independent of any energy model.

Subchronic Toxicity of the New Iodine Complex in Dogs and Rats.

Background: Complexes of iodine (povidone-iodine and cadexomers) are among the most important antiseptics used in clinical and veterinary medicines. However, high local irritation activity and systemic toxicity limits their oral administration. The purpose of the study was to compare the effect of a new complex of iodine (PA, potentiator of anticancer antibiotics), in which iodine is coordinated by carbohydrates and polypeptides) on the organisms of rats and dogs treated orally with the drug for 30 days. Methods: Wistar rats and Beagle dogs served as experimental animal models. Effect of PA on the animal organism was examined through the measurements of hormones level changes, hematological and clinical chemistry parameters alterations, necropsy and histological examination. Results: The established maximum tolerated dose (MTD) of 2,000 mg/kg PA led to a decrease in the rate of body weight gain in male and female rats. Changes in hematological and certain biochemical parameters in rats at doses of 1,000 and 2,000 mg/kg were observed. Histological study of the thyroid gland revealed changes in the shape and size of the follicles along with colloid resorption. Administration of a half of MTD (180 mg/kg) and lower doses did not result in any change in dogs (thyroid-stimulating hormone, triiodothyronine, and thyroxine). Conclusions: The results of our study show that the pathogenetic action of PA takes place along the path of induction of an inflammatory response with the development of thyrotoxicosis, rather than hypothyroidism. The mechanism of induction of an inflammatory response is also confirmed by histological studies of lesions of the thyroid gland and testes in rats (Figure S1). The no-observed-adverse-effect level (NOAEL) of PA is estimated to be 180 mg/kg (or iodine 22.8 mg/kg) in dogs, which is equivalent to 100 mg/kg (or iodine 12.3 mg/kg) in humans.

Genotoxic activity of the new pharmaceutical FS-1 in Salmonella/microsome test and mouse lymphoma L5178Y cells.

PURPOSE: The purpose of this study was to determine possible genotoxic effects of a new very promising antibacterial/antiviral drug FS-1. METHODS: The drug was tested in TA98, TA100, TA102, TA1535 and TA1537 strains of Salmonella (Ames test) with and without metabolic activation, and also in mouse lymphoma L5178Y cells by means of micronucleus and comet assays. In microbes the drug was tested at concentrations up to 500 µg/plate and in mouse lymphoma cells up to 2,000 µg/ml. RESULTS: In both test-systems in all experiments completely negative results were obtained although FS-1 was tested at maximum tolerated doses. CONCLUSIONS: The drug is not genotoxic. This is advantageous because many antibacterial/antiviral drugs possess such activity.

Parasporins as new natural anticancer agents: a review.

In 1999 Mizuki and co-authors studied for the first time the parasporal inclusion proteins extracted from B. thuringiensis strains (a Gram-positive, soil-dwelling bacterium) for cytotoxic activity against human leukaemia T-cells. Later some other proteins with this unusual property to recognize human leukemic cells were isolated from this strain of bacteria and named parasporins. At present 6 types of parasporins are identified and characterized. This review summarizes the properties of these new potentially useful antitumor agents of natural origin. Various types of parasporins possess unique cytotoxic mechanisms against cancer cells. The cytotoxic activity for cancer cells makes parasporins possible candidates for anticancer agents in clinical oncology. Recently, genetic engineering was applied for the production of parasporins and the gene responsible for the production of the proteins was expressed in E. coli. However, there are virtually no data regarding the cytotoxic (antitumor) activity of parasporins in vivo. These relatively new cytotoxic proteins warrant further investigation, especially in rodents, for possible application in clinical oncology.

Enhanced gradient for training restricted Boltzmann machines.

Restricted Boltzmann machines (RBMs) are often used as building blocks in greedy learning of deep networks. However, training this simple model can be laborious. Traditional learning algorithms often converge only with the right choice of metaparameters that specify, for example, learning rate scheduling and the scale of the initial weights. They are also sensitive to specific data representation. An equivalent RBM can be obtained by flipping some bits and changing the weights and biases accordingly, but traditional learning rules are not invariant to such transformations. Without careful tuning of these training settings, traditional algorithms can easily get stuck or even diverge. In this letter, we present an enhanced gradient that is derived to be invariant to bit-flipping transformations. We experimentally show that the enhanced gradient yields more stable training of RBMs both when used with a fixed learning rate and an adaptive one.

Vascular endothelial growth factor in anterior chamber liquid patients with diabetic retinopathy, cataract and neovascular glaucoma.

INTRODUCTION: The aims of this study were: (1) to investigate the association of vascular endothelial growth factor isoform A (VEGF-A) concentration in the anterior chamber liquid (ACL) with vascular proliferation in patients with diabetic retinopathy (DR) who had undergone surgical treatment for cataract and neovascular glaucoma; (2) to analyze the association of VEGF-A level in ACL with the cataract surgery outcomes. MATERIALS AND METHODS: Undiluted aqueous fluid samples were obtained from 207 eyes of patients who underwent intraocular surgery, 136 patients with diabetes mellitus (DM) and 22 patients without DM. The ACL samples were obtained during operation. The VEGF-A levels were analyzed by enzyme-linked immunosorbent assay. RESULTS: The lowest VEGF-A levels were in diabetic patients without signs of DR [22.75 pg/mL (10.78; 63.36)]. More severe DR tended to occur in diabetic patients with higher VEGF-A levels in ACL. In diabetic patients with proliferative DR (PDR), VEGF-A levels were significantly higher [336.6 pg/mL (232.3; 410.74)] than in patients without DR P < 0.0001. In patients with terminal stage of DR [neovascular glaucoma (NG)], VEGF-A levels were dramatically higher and attained 1,634.01 pg/mL (610.69; 2657.33). In non-diabetic patients, VEGF-A levels were 95.07 pg/ml (60.92; 129.22). The best visual acuity results in post-operative period were observed in the group of diabetic patients without DR. In the group of patients with PDR, post-operative visual acuity [0.26 (0.1; 0.42)] was similar to visual acuity before operation [0.29 (0.13; 0.44)]. There was no significant increase in visual acuity due to cataract surgery. In 52.4% patients, no complications had occurred by the end of the follow-up period. In 40% patients, retinal laser coagulation was performed, and in 7.6% patients NG had developed. CONCLUSION: VEGF-A level in ACL increases with DR progression and may be of prognostic value in evaluating the potential risk of further neovascularization progression in diabetic patients.

Exploratory analysis of climate data using source separation methods.

We present an example of exploratory data analysis of climate measurements using a recently developed denoising source separation (DSS) framework. We analyzed a combined dataset containing daily measurements of three variables: surface temperature, sea level pressure and precipitation around the globe, for a period of 56 years. Components exhibiting slow temporal behavior were extracted using DSS with linear denoising. The first component, most prominent in the interannual time scale, captured the well-known El Niño-Southern Oscillation (ENSO) phenomenon and the second component was close to the derivative of the first one. The slow components extracted in a wider frequency range were further rotated using a frequency-based separation criterion implemented by DSS with nonlinear denoising. The rotated sources give a meaningful representation of the slow climate variability as a combination of trends, interannual oscillations, the annual cycle and slowly changing seasonal variations. Again, components related to the ENSO phenomenon emerge very clearly among the found sources.

Nonlinear dynamical factor analysis for state change detection.

Changes in a dynamical process are often detected by monitoring selected indicators directly obtained from the process observations, such as the mean values or variances. Standard change detection algorithms such as the Shewhart control charts or the cumulative sum (CUSUM) algorithm are often based on such first- and second-order statistics. Much better results can be obtained if the dynamical process is properly modeled, for example by a nonlinear state-space model, and then the accuracy of the model is monitored over time. The success of the latter approach depends largely on the quality of the model. In practical applications like industrial processes, the state variables, dynamics, and observation mapping are rarely known accurately. Learning from data must be used; however, methods for the simultaneous estimation of the state and the unknown nonlinear mappings are very limited. We use a novel method of learning a nonlinear state-space model, the nonlinear dynamical factor analysis (NDFA) algorithm. It takes a set of multivariate observations over time and fits blindly a generative dynamical latent variable model, resembling nonlinear independent component analysis. We compare the performance of the model in process change detection to various traditional methods. It is shown that NDFA outperforms the classical methods by a wide margin in a variety of cases where the underlying process dynamics changes.