DFTB+, a software package for efficient approximate density functional theory based atomistic simulations.

DFTB+ is a versatile community developed open source software package offering fast and efficient methods for carrying out atomistic quantum mechanical simulations. By implementing various methods approximating density functional theory (DFT), such as the density functional based tight binding (DFTB) and the extended tight binding method, it enables simulations of large systems and long timescales with reasonable accuracy while being considerably faster for typical simulations than the respective ab initio methods. Based on the DFTB framework, it additionally offers approximated versions of various DFT extensions including hybrid functionals, time dependent formalism for treating excited systems, electron transport using non-equilibrium Green's functions, and many more. DFTB+ can be used as a user-friendly standalone application in addition to being embedded into other software packages as a library or acting as a calculation-server accessed by socket communication. We give an overview of the recently developed capabilities of the DFTB+ code, demonstrating with a few use case examples, discuss the strengths and weaknesses of the various features, and also discuss on-going developments and possible future perspectives.

[The comparative study of the early stages of sensory and sensory-motor information processing in aggressive and non-aggressive schizophrenic patients]. / Sravnitel'naia otsenka pokazatelei rannikh étapov obrabotki sensornoi i sensomotornoi informatsii u agressivnykh i neagressivnykh bol'nykh shizofreniei.

AIM: To study the early stages of sensory (P50 suppression) and sensorimotor (prepulse modification of acoustic startle reaction - ASR) information processing in aggressive and non-aggressive patients with schizophrenia. MATERIAL AND METHODS: Two groups (aggressive group A, 44 s. and non-aggressive group NA, 27 s.) and 48 healthy people were studied using oculography and encephalography. Symptoms of schizophrenia were assessed by PANSS. RESULTS AND CONCLUSION: The analysis of ASR and P50 revealed the increased ASR latency, deficit of prepulse inhibition (PPI) with 60 ms leading interval, and P50 suppression in patients compared to controls. Aggressive behavior impacted on lateral features of ASR changes in the group A. The increase of ASR amplitude and the decline of prepulse facilitation were found in the group NA. The P50 analysis revealed the increased P50 amplitude and the shortened latency in response to 2nd stimulus in the pair. The increase in scores on item G14 of the PANSS (Poor impulse control) in the group A and its correlations with ASR and P50 parameters in this group suggested the important role of impulsivity in aggressive behavior in the studied sample of aggressive patients. ASR and P50 characteristics in the group NA are associated with increased sensitivity to incoming stimuli and such personality traits as increased alertness and fear, which might inhibit aggression. Thus, indicators of the early stages of information processing can be considered as potential predictors of aggressive behavior.

Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions.

Machine learning advances chemistry and materials science by enabling large-scale exploration of chemical space based on quantum chemical calculations. While these models supply fast and accurate predictions of atomistic chemical properties, they do not explicitly capture the electronic degrees of freedom of a molecule, which limits their applicability for reactive chemistry and chemical analysis. Here we present a deep learning framework for the prediction of the quantum mechanical wavefunction in a local basis of atomic orbitals from which all other ground-state properties can be derived. This approach retains full access to the electronic structure via the wavefunction at force-field-like efficiency and captures quantum mechanics in an analytically differentiable representation. On several examples, we demonstrate that this opens promising avenues to perform inverse design of molecular structures for targeting electronic property optimisation and a clear path towards increased synergy of machine learning and quantum chemistry.

SchNetPack: A Deep Learning Toolbox For Atomistic Systems.

SchNetPack is a toolbox for the development and application of deep neural networks that predict potential energy surfaces and other quantum-chemical properties of molecules and materials. It contains basic building blocks of atomistic neural networks, manages their training, and provides simple access to common benchmark datasets. This allows for an easy implementation and evaluation of new models. For now, SchNetPack includes implementations of (weighted) atom-centered symmetry functions and the deep tensor neural network SchNet, as well as ready-to-use scripts that allow one to train these models on molecule and material datasets. Based on the PyTorch deep learning framework, SchNetPack allows one to efficiently apply the neural networks to large datasets with millions of reference calculations, as well as parallelize the model across multiple GPUs. Finally, SchNetPack provides an interface to the Atomic Simulation Environment in order to make trained models easily accessible to researchers that are not yet familiar with neural networks.

SchNet - A deep learning architecture for molecules and materials.

Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Analysis of altered gene expression in rat soleus muscle atrophied by disuse.

The present study involved a global analysis of genes whose expression was modified in rat soleus muscle atrophied after hindlimb suspension (HS). HS muscle unloading is a common model for muscle disuse that especially affects antigravity slow-twitch muscles such as the soleus muscle. A cDNA cloning strategy, based on suppression subtractive hybridization technology, led to the construction of two normalized soleus muscle cDNA libraries that were subtracted in opposite directions, i.e., atrophied soleus muscle cDNAs subtracted by control cDNAs and vice versa. Differential screening of the two libraries revealed 34 genes with altered expression in HS soleus muscle, including 11 novel cDNAs, in addition to the 2X and 2B myosin heavy chain genes expressed only in soleus muscles after HS. Gene up- and down-regulations were quantified by reverse Northern blot and classical Northern blot analysis. The 25 genes with known functions fell into seven important functional categories. The homogeneity of gene alterations within each category gave several clues for unraveling the interplay of cellular events implied in the muscle atrophy phenotype. In particular, our results indicate that modulations in slow- and fast-twitch-muscle component balance, the protein synthesis/secretion pathway, and the extracellular matrix/cytoskeleton axis are likely to be key molecular mechanisms of muscle atrophy. In addition, the cloning of novel cDNAs underlined the efficiency of the chosen technical approach and gave novel possibilities to further decipher the molecular mechanisms of muscle atrophy.

Overexpression of Hoxc13 in differentiating keratinocytes results in downregulation of a novel hair keratin gene cluster and alopecia.

Studying the roles of Hox genes in normal and pathological development of skin and hair requires identification of downstream target genes in genetically defined animal models. We show that transgenic mice overexpressing Hoxc13 in differentiating keratinocytes of hair follicles develop alopecia, accompanied by a progressive pathological skin condition that resembles ichthyosis. Large-scale analysis of differential gene expression in postnatal skin of these mice identified 16 previously unknown and 13 known genes as presumptive Hoxc13 targets. The majority of these targets are downregulated and belong to a subgroup of genes that encode hair-specific keratin-associated proteins (KAPs). Genomic mapping using a mouse hamster radiation hybrid panel showed these genes to reside in a novel KAP gene cluster on mouse chromosome 16 in a region of conserved linkage with human chromosome 21q22.11. Furthermore, data obtained by Hoxc13/lacZ reporter gene analysis in mice that overexpress Hoxc13 suggest negative autoregulatory feedback control of Hoxc13 expression levels, thus providing an entry point for elucidating currently unknown mechanisms that are required for regulating quantitative levels of Hox gene expression. Combined, these results provide a framework for understanding molecular mechanisms of Hoxc13 function in hair growth and development.

Identification of altered gene expression in skeletal muscles from Duchenne muscular dystrophy patients.

Mutations in the dystrophin gene lead to dystrophin deficiency, which is the cause of Duchenne muscular dystrophy (DMD). This important discovery more than 10 years ago opened a new field for very productive investigations. However, the exact functions of dystrophin are still not fully understood and the complex process leading to subsequent muscle fiber necrosis has not been clearly described; hence there has not yet been any marked improvement in patient treatment. To decipher the molecular mechanisms induced by a lack of dystrophin, we started identifying genes whose expression is altered in DMD skeletal muscles. The approach was based on differential screening of a human muscle cDNA array. Nine genes were found to be up- or downregulated. Our results indicate expression alterations in mitochondrial genes, titin, a muscle transcription factor and three novel genes. First characterizations of these novel genes indicated that two of them have striated muscle tissue specificity.

Large-scale analysis of differential gene expression in the hindlimb muscles and diaphragm of mdx mouse.

The mdx mouse is an animal model for Duchenne muscular dystrophy (DMD), which is caused by the absence of dystrophin. Mdx limb muscles substantially compensate for the lack of dystrophin while the diaphragm is affected like DMD skeletal muscles. To understand better the complex cascade of molecular events leading to muscle degeneration and compensatory processes in mdx muscles, we analyzed alterations of gene expression in mdx hindlimb and diaphragm muscles as compared to their normal counterparts. The strategy was based on suppression subtractive hybridization followed by reverse Northern quantitative hybridization. Four subtracted/normalized libraries, containing cDNA clones up- or downregulated in mdx hindlimb muscles or diaphragm, were constructed and a total of 1536 cDNA clones were analyzed. Ninety-three cDNAs were found to be differentially expressed in mdx hindlimb muscles and/or diaphragm. They corresponded to 54 known genes and 39 novel cDNAs. The potential role of the known genes is discussed in the context of the mdx phenotype.