Search

ClusterCAD is a web-based toolkit designed to leverage the collinear structure and deterministic logic of type I modular polyketide synthases (PKSs) for synthetic biology applications. The unique organization of these megasynthases, combined with the diversity of their catalytic domain building blocks, has fueled an interest in harnessing the biosynthetic potential of PKSs for the microbial production of both novel natural product analogs and industrially relevant small molecules. However, a limited theoretical understanding of the determinants of PKS fold and function poses a substantial barrier to the design of active variants, and identifying strategies to reliably construct functional PKS chimeras remains an active area of research. In this work, we formalize a paradigm for the design of PKS chimeras and introduce ClusterCAD as a computational platform to streamline and simplify the process of designing experiments to test strategies for engineering PKS variants. ClusterCAD provides chemical structures with stereochemistry for the intermediates generated by each PKS module, as well as sequence- and structure-based search tools that allow users to identify modules based either on amino acid sequence or on the chemical structure of the cognate polyketide intermediate. ClusterCAD can be accessed at https://clustercad.jbei.org and at http://clustercad.igb.uci.edu.

Genome biology approaches have made enormous contributions to our understanding of biological rhythms, particularly in identifying outputs of the clock, including RNAs, proteins, and metabolites, whose abundance oscillates throughout the day. These methods hold significant promise for future discovery, particularly when combined with computational modeling. However, genome-scale experiments are costly and laborious, yielding "big data" that are conceptually and statistically difficult to analyze. There is no obvious consensus regarding design or analysis. Here we discuss the relevant technical considerations to generate reproducible, statistically sound, and broadly useful genome-scale data. Rather than suggest a set of rigid rules, we aim to codify principles by which investigators, reviewers, and readers of the primary literature can evaluate the suitability of different experimental designs for measuring different aspects of biological rhythms. We introduce CircaInSilico, a web-based application for generating synthetic genome biology data to benchmark statistical methods for studying biological rhythms. Finally, we discuss several unmet analytical needs, including applications to clinical medicine, and suggest productive avenues to address them.

Spinal metastasis, a metastatic cancer of the spine, is the most common malignant disease in the spine. In this study, we investigate the feasibility of automated spinal metastasis detection in magnetic resonance imaging (MRI) by using deep learning methods. To accommodate the large variability in metastatic lesion sizes, we develop a Siamese deep neural network approach comprising three identical subnetworks for multi-resolution analysis and detection of spinal metastasis. At each location of interest, three image patches at three different resolutions are extracted and used as the input to the networks. To further reduce the false positives (FPs), we leverage the similarity between neighboring MRI slices, and adopt a weighted averaging strategy to aggregate the results obtained by the Siamese neural networks. The detection performance is evaluated on a set of 26 cases using a free-response receiver operating characteristic (FROC) analysis. The results show that the proposed approach correctly detects all the spinal metastatic lesions while producing only 0.40 FPs per case. At a true positive (TP) rate of 90%, the use of the aggregation reduces the FPs from 0.375 FPs per case to 0.207 FPs per case, a nearly 44.8% reduction. The results indicate that the proposed Siamese neural network method, combined with the aggregation strategy, provide a viable strategy for the automated detection of spinal metastasis in MRI images.

Coronary artery disease is a major cause of death in women. Breast arterial calcifications (BACs), detected inmammograms, can be useful riskmarkers associated with the disease. We investigate the feasibility of automated and accurate detection ofBACsinmammograms for risk assessment of coronary artery disease. We develop a 12-layer convolutional neural network to discriminate BAC from non-BAC and apply a pixelwise, patch-based procedure for BAC detection. To assess the performance of the system, we conduct a reader study to provide ground-truth information using the consensus of human expert radiologists. We evaluate the performance using a set of 840 full-field digital mammograms from 210 cases, using both free-responsereceiveroperatingcharacteristic (FROC) analysis and calcium mass quantification analysis. The FROC analysis shows that the deep learning approach achieves a level of detection similar to the human experts. The calcium mass quantification analysis shows that the inferred calcium mass is close to the ground truth, with a linear regression between them yielding a coefficient of determination of 96.24%. Taken together, these results suggest that deep learning can be used effectively to develop an automated system for BAC detection inmammograms to help identify and assess patients with cardiovascular risks.

Recent human exome-sequencing studies have implicated polymorphic Brg1-associated factor (BAF) complexes (mammalian SWI/SNF chromatin remodeling complexes) in several intellectual disabilities and cognitive disorders, including autism. However, it remains unclear how mutations in BAF complexes result in impaired cognitive function. Post-mitotic neurons express a neuron-specific assembly, nBAF, characterized by the neuron-specific subunit BAF53b. Subdomain 2 of BAF53b is essential for the differentiation of neuronal precursor cells into neurons. We generated transgenic mice lacking subdomain 2 of Baf53b (BAF53bΔSB2). Long-term synaptic potentiation (LTP) and long-term memory, both of which are associated with phosphorylation of the actin severing protein cofilin, were assessed in these animals. A phosphorylation mimic of cofilin was stereotaxically delivered into the hippocampus of BAF53bΔSB2 mice in an effort to rescue LTP and memory. BAF53bΔSB2 mutant mice show impairments in phosphorylation of synaptic cofilin, LTP, and memory. Both the synaptic plasticity and memory deficits are rescued by overexpression of a phosphorylation mimetic of cofilin. Baseline physiology and behavior were not affected by the mutation or the experimental treatment. This study suggests a potential link between nBAF function, actin cytoskeletal remodeling at the dendritic spine, and memory formation. This work shows that a targeted manipulation of synaptic function can rescue adult plasticity and memory deficits caused by manipulations of nBAF, and thereby provides potential novel avenues for therapeutic development for multiple intellectual disability disorders.

MicroRNAs (miRNAs) are known to mediate post-transcriptional gene regulation, but their role in postnatal brain development is still poorly explored. We show that the expression of many miRNAs is dramatically regulated during functional maturation of the mouse visual cortex with miR-132/212 family being one of the top upregulated miRNAs. Age-downregulated transcripts are significantly enriched in miR-132/miR-212 putative targets and in genes upregulated in miR-132/212 null mice. At a functional level, miR-132/212 deletion affects development of receptive fields of cortical neurons determining a specific impairment of binocular matching of orientation preference, but leaving orientation and direction selectivity unaltered. This deficit is associated with reduced depth perception in the visual cliff test. Deletion of miR-132/212 from forebrain excitatory neurons replicates the binocular matching deficits. Thus, miR-132/212 family shapes the age-dependent transcriptome of the visual cortex during a specific developmental window resulting in maturation of binocular cortical cells and depth perception.

BackgroundThe majority of cancer cells have a unique metabolic requirement for methionine that is not observed in normal, non-tumorigenic cells. This phenotype is described as "methionine dependence" or "methionine stress sensitivity" in which cancer cells are unable to proliferate when methionine has been replaced with its metabolic precursor, homocysteine, in cell culture growth media. We focus on the metabolic response to methionine stress in the triple negative breast cancer cell line MDA-MB-468 and its methionine insensitive derivative cell line MDA-MB-468res-R8.ResultsUsing a variety of techniques including fluorescence lifetime imaging microscopy (FLIM) and extracellular flux assays, we identified a metabolic down-regulation of oxidative phosphorylation in both MDA-MB-468 and MDA-MB-468res-R8 cell types when cultured in homocysteine media. Untargeted metabolomics was performed by way of gas chromatography/time-of-flight mass spectrometry on both cell types cultured in homocysteine media over a period of 2 to 24 h. We determined unique metabolic responses between the two cell lines in specific pathways including methionine salvage, purine/pyrimidine synthesis, and the tricarboxylic acid cycle. Stable isotope tracer studies using deuterium-labeled homocysteine indicated a redirection of homocysteine metabolism toward the transsulfuration pathway and glutathione synthesis. This data corroborates with increased glutathione levels concomitant with increased levels of oxidized glutathione. Redirection of homocysteine flux resulted in reduced generation of methionine from homocysteine particularly in MDA-MB-468 cells. Consequently, synthesis of the important one-carbon donor S-adenosylmethionine (SAM) was decreased, perturbing the SAM to S-adenosylhomocysteine ratio in MDA-MB-468 cells, which is an indicator of the cellular methylation potential.ConclusionThis study indicates a differential metabolic response between the methionine sensitive MDA-MB-468 cells and the methionine insensitive derivative cell line MDA-MB-468res-R8. Both cell lines appear to experience oxidative stress when methionine was replaced with its metabolic precursor homocysteine, forcing cells to redirect homocysteine metabolism toward the transsulfuration pathway to increase glutathione synthesis. The methionine stress resistant MDA-MB-468res-R8 cells responded to this cellular stress earlier than the methionine stress sensitive MDA-MB468 cells and coped better with metabolic demands. Additionally, it is evident that S-adenosylmethionine metabolism is dependent on methionine availability in cancer cells, which cannot be sufficiently supplied by homocysteine metabolism under these conditions.

The liver circadian clock is reprogrammed by nutritional challenge through the rewiring of specific transcriptional pathways. As the gut microbiota is tightly connected to host metabolism, whose coordination is governed by the circadian clock, we explored whether gut microbes influence circadian homeostasis and how they distally control the peripheral clock in the liver. Using fecal transplant procedures we reveal that, in response to high-fat diet, the gut microbiota drives PPARγ-mediated activation of newly oscillatory transcriptional programs in the liver. Moreover, antibiotics treatment prevents PPARγ-driven transcription in the liver, underscoring the essential role of gut microbes in clock reprogramming and hepatic circadian homeostasis. Thus, a specific molecular signature characterizes the influence of the gut microbiome in the liver, leading to the transcriptional rewiring of hepatic metabolism.

MotivationNot only sequence data continue to outpace annotation information, but also the problem is further exacerbated when organisms are underrepresented in the annotation databases. This is the case with non-human-pathogenic viruses which occur frequently in metagenomic projects. Thus, there is a need for tools capable of detecting and classifying viral sequences.ResultsWe describe VIRALpro a new effective tool for identifying capsid and tail protein sequences, which are the cornerstones toward viral sequence annotation and viral genome classification.Availability and implementationThe data, software and corresponding web server are available from http://scratch.proteomics.ics.uci.edu as part of the SCRATCH suite.Contactclovis.galiez@inria.fr or pfbaldi@uci.eduSupplementary informationSupplementary data are available at Bioinformatics online.

The circadian clock controls metabolic and physiological processes through finely tuned molecular mechanisms. The clock is remarkably plastic and adapts to exogenous "zeitgebers," such as light and nutrition. How a pathological condition in a given tissue influences systemic circadian homeostasis in other tissues remains an unanswered question of conceptual and biomedical importance. Here, we show that lung adenocarcinoma operates as an endogenous reorganizer of circadian metabolism. High-throughput transcriptomics and metabolomics revealed unique signatures of transcripts and metabolites cycling exclusively in livers of tumor-bearing mice. Remarkably, lung cancer has no effect on the core clock but rather reprograms hepatic metabolism through altered pro-inflammatory response via the STAT3-Socs3 pathway. This results in disruption of AKT, AMPK, and SREBP signaling, leading to altered insulin, glucose, and lipid metabolism. Thus, lung adenocarcinoma functions as a potent endogenous circadian organizer (ECO), which rewires the pathophysiological dimension of a distal tissue such as the liver. PAPERCLIP.

Diagnosis and therapeutic interventions in pathological conditions rely upon clinical monitoring of key metabolites in the serum. Recent studies show that a wide range of metabolic pathways are controlled by circadian rhythms whose oscillation is affected by nutritional challenges, underscoring the importance of assessing a temporal window for clinical testing and thereby questioning the accuracy of the reading of critical pathological markers in circulation. We have been interested in studying the communication between peripheral tissues under metabolic homeostasis perturbation. Here we present a comparative circadian metabolomic analysis on serum and liver in mice under high fat diet. Our data reveal that the nutritional challenge induces a loss of serum metabolite rhythmicity compared with liver, indicating a circadian misalignment between the tissues analyzed. Importantly, our results show that the levels of serum metabolites do not reflect the circadian liver metabolic signature or the effect of nutritional challenge. This notion reveals the possibility that misleading reads of metabolites in circulation may result in misdiagnosis and improper treatments. Our findings also demonstrate a tissue-specific and time-dependent disruption of metabolic homeostasis in response to altered nutrition.

Yarrowia lipolytica, an oleaginous yeast, is capable of accumulating significant cellular mass in lipid making it an important source of biosustainable hydrocarbon-based chemicals. In spite of a similar number of protein-coding genes to that in other Hemiascomycetes, the Y. lipolytica genome is almost double that of model yeasts. Despite its economic importance and several distinct strains in common use, an independent genome assembly exists for only one strain. We report here a de novo annotated assembly of the chromosomal genome of an industrially-relevant strain, W29/CLIB89, determined by hybrid next-generation sequencing. For the first time, each Y. lipolytica chromosome is represented by a single contig. The telomeric rDNA repeats were localized by Irys long-range genome mapping and one complete copy of the rDNA sequence is reported. Two large structural variants and retroelement differences with reference strain CLIB122 including a full-length, novel Ty3/Gypsy long terminal repeat (LTR) retrotransposon and multiple LTR-like sequences are described. Strikingly, several of these are adjacent to RNA polymerase III-transcribed genes, which are almost double in number in Y. lipolytica compared to other Hemiascomycetes. In addition to previously-reported dimeric RNA polymerase III-transcribed genes, tRNA pseudogenes were identified. Multiple full-length and truncated LINE elements are also present. Therefore, although identified transposons do not constitute a significant fraction of the Y. lipolytica genome, they could have played an active role in its evolution. Differences between the sequence of this strain and of the existing reference strain underscore the utility of an additional independent genome assembly for this economically important organism.

BackgroundA number of algorithms have been proposed to predict the biological targets of diverse molecules. Some are structure-based, but the most common are ligand-based and use chemical fingerprints and the notion of chemical similarity. These methods tend to be computationally faster than others, making them particularly attractive tools as the amount of available data grows.ResultsUsing a ChEMBL-derived database covering 490,760 molecule-protein interactions and 3236 protein targets, we conduct a large-scale assessment of the performance of several target-prediction algorithms at predicting drug-target activity. We assess algorithm performance using three validation procedures: standard tenfold cross-validation, tenfold cross-validation in a simulated screen that includes random inactive molecules, and validation on an external test set composed of molecules not present in our database.ConclusionsWe present two improvements over current practice. First, using a modified version of the influence-relevance voter (IRV), we show that using molecule potency data can improve target prediction. Second, we demonstrate that random inactive molecules added during training can boost the accuracy of several algorithms in realistic target-prediction experiments. Our potency-sensitive version of the IRV (PS-IRV) obtains the best results on large test sets in most of the experiments. Models and software are publicly accessible through the chemoinformatics portal at http://chemdb.ics.uci.edu/.

Candida albicans is associated with humans as both a harmless commensal organism and a pathogen. Cph2 is a transcription factor whose DNA binding domain is similar to that of mammalian sterol response element binding proteins (SREBPs). SREBPs are master regulators of cellular cholesterol levels and are highly conserved from fungi to mammals. However, ergosterol biosynthesis is regulated by the zinc finger transcription factor Upc2 in C. albicans and several other yeasts. Cph2 is not necessary for ergosterol biosynthesis but is important for colonization in the murine gastrointestinal (GI) tract. Here we demonstrate that Cph2 is a membrane-associated transcription factor that is processed to release the N-terminal DNA binding domain like SREBPs, but its cleavage is not regulated by cellular levels of ergosterol or oxygen. Chromatin immunoprecipitation sequencing (ChIP-seq) shows that Cph2 binds to the promoters of HMS1 and other components of the regulatory circuit for GI tract colonization. In addition, 50% of Cph2 targets are also bound by Hms1 and other factors of the regulatory circuit. Several common targets function at the head of the glycolysis pathway. Thus, Cph2 is an integral part of the regulatory circuit for GI colonization that regulates glycolytic flux. Transcriptome sequencing (RNA-seq) shows a significant overlap in genes differentially regulated by Cph2 and hypoxia, and Cph2 is important for optimal expression of some hypoxia-responsive genes in glycolysis and the citric acid cycle. We suggest that Cph2 and Upc2 regulate hypoxia-responsive expression in different pathways, consistent with a synthetic lethal defect of the cph2 upc2 double mutant in hypoxia.

MotivationCircadian oscillations have been observed in animals, plants, fungi and cyanobacteria and play a fundamental role in coordinating the homeostasis and behavior of biological systems. Genetically encoded molecular clocks found in nearly every cell, based on negative transcription/translation feedback loops and involving only a dozen genes, play a central role in maintaining these oscillations. However, high-throughput gene expression experiments reveal that in a typical tissue, a much larger fraction ([Formula: see text]) of all transcripts oscillate with the day-night cycle and the oscillating species vary with tissue type suggesting that perhaps a much larger fraction of all transcripts, and perhaps also other molecular species, may bear the potential for circadian oscillations.ResultsTo better quantify the pervasiveness and plasticity of circadian oscillations, we conduct the first large-scale analysis aggregating the results of 18 circadian transcriptomic studies and 10 circadian metabolomic studies conducted in mice using different tissues and under different conditions. We find that over half of protein coding genes in the cell can produce transcripts that are circadian in at least one set of conditions and similarly for measured metabolites. Genetic or environmental perturbations can disrupt existing oscillations by changing their amplitudes and phases, suppressing them or giving rise to novel circadian oscillations. The oscillating species and their oscillations provide a characteristic signature of the physiological state of the corresponding cell/tissue. Molecular networks comprise many oscillator loops that have been sculpted by evolution over two trillion day-night cycles to have intrinsic circadian frequency. These oscillating loops are coupled by shared nodes in a large network of coupled circadian oscillators where the clock genes form a major hub. Cells can program and re-program their circadian repertoire through epigenetic and other mechanisms.Availability and implementationHigh-resolution and tissue/condition specific circadian data and networks available at http://circadiomics.igb.uci.edu.Contactpfbaldi@ics.uci.eduSupplementary informationSupplementary data are available at Bioinformatics online.

© 2015 Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved. Despite an extensive body of reported information about peripheral and central mechanisms involved in the pathophysiology of IBS symptoms, no comprehensive disease model has emerged that would guide the development of novel, effective therapies. In this Review, we will first describe novel insights into some key components of brain–gut interactions, starting with the emerging findings of distinct functional and structural brain signatures of IBS. We will then point out emerging correlations between these brain networks and genomic, gastrointestinal, immune and gut-microbiome-related parameters. We will incorporate this new information, as well as the reported extensive literature on various peripheral mechanisms, into a systems-based disease model of IBS, and discuss the implications of such a model for improved understanding of the disorder, and for the development of more-effective treatment approaches in the future.

A considerable body of evidence supports the role of mitochondrial dysfunction in psychiatric disorders and mitochondrial DNA (mtDNA) mutations are known to alter brain energy metabolism, neurotransmission, and cause neurodegenerative disorders. Genetic studies focusing on common nuclear genome variants associated with these disorders have produced genome wide significant results but those studies have not directly studied mtDNA variants. The purpose of this study is to investigate, using next generation sequencing, the involvement of mtDNA variation in bipolar disorder, schizophrenia, major depressive disorder, and methamphetamine use. MtDNA extracted from multiple brain regions and blood were sequenced (121 mtDNA samples with an average of 8,800x coverage) and compared to an electronic database containing 26,850 mtDNA genomes. We confirmed novel and rare variants, and confirmed next generation sequencing error hotspots by traditional sequencing and genotyping methods. We observed a significant increase of non-synonymous mutations found in individuals with schizophrenia. Novel and rare non-synonymous mutations were found in psychiatric cases in mtDNA genes: ND6, ATP6, CYTB, and ND2. We also observed mtDNA heteroplasmy in brain at a locus previously associated with schizophrenia (T16519C). Large differences in heteroplasmy levels across brain regions within subjects suggest that somatic mutations accumulate differentially in brain regions. Finally, multiplasmy, a heteroplasmic measure of repeat length, was observed in brain from selective cases at a higher frequency than controls. These results offer support for increased rates of mtDNA substitutions in schizophrenia shown in our prior results. The variable levels of heteroplasmic/multiplasmic somatic mutations that occur in brain may be indicators of genetic instability in mtDNA.

LEF/TCFs direct the final step in Wnt/β-catenin signalling by recruiting β-catenin to genes for activation of transcription. Ancient, non-vertebrate TCFs contain two DNA binding domains, a High Mobility Group box for recognition of the Wnt Response Element (WRE; 5'-CTTTGWWS-3') and the C-clamp domain for recognition of the GC-rich Helper motif (5'-RCCGCC-3'). Two vertebrate TCFs (TCF-1/TCF7 and TCF-4/TCF7L2) use the C-clamp as an alternatively spliced domain to regulate cell-cycle progression, but how the C-clamp influences TCF binding and activity genome-wide is not known. Here, we used a doxycycline inducible system with ChIP-seq to assess how the C-clamp influences human TCF1 binding genome-wide. Metabolic pulse-labeling of nascent RNA with 4'Thiouridine was used with RNA-seq to connect binding to the Wnt transcriptome. We find that the C-clamp enables targeting to a greater number of gene loci for stronger occupancy and transcription regulation. The C-clamp uses Helper sites concurrently with WREs for gene targeting, but it also targets TCF1 to sites that do not have readily identifiable canonical WREs. The coupled ChIP-seq/4'Thiouridine-seq analysis identified new Wnt target genes, including additional regulators of cell proliferation. Thus, C-clamp containing isoforms of TCFs are potent transcriptional regulators with an expanded transcriptome directed by C-clamp-Helper site interactions.

[This corrects the article DOI: 10.1016/j.molmet.2013.10.005.].

MotivationAccurately predicting protein secondary structure and relative solvent accessibility is important for the study of protein evolution, structure and function and as a component of protein 3D structure prediction pipelines. Most predictors use a combination of machine learning and profiles, and thus must be retrained and assessed periodically as the number of available protein sequences and structures continues to grow.ResultsWe present newly trained modular versions of the SSpro and ACCpro predictors of secondary structure and relative solvent accessibility together with their multi-class variants SSpro8 and ACCpro20. We introduce a sharp distinction between the use of sequence similarity alone, typically in the form of sequence profiles at the input level, and the additional use of sequence-based structural similarity, which uses similarity to sequences in the Protein Data Bank to infer annotations at the output level, and study their relative contributions to modern predictors. Using sequence similarity alone, SSpro's accuracy is between 79 and 80% (79% for ACCpro) and no other predictor seems to exceed 82%. However, when sequence-based structural similarity is added, the accuracy of SSpro rises to 92.9% (90% for ACCpro). Thus, by combining both approaches, these problems appear now to be essentially solved, as an accuracy of 100% cannot be expected for several well-known reasons. These results point also to several open technical challenges, including (i) achieving on the order of ≥ 80% accuracy, without using any similarity with known proteins and (ii) achieving on the order of ≥ 85% accuracy, using sequence similarity alone.Availability and implementationSSpro, SSpro8, ACCpro and ACCpro20 programs, data and web servers are available through the SCRATCH suite of protein structure predictors at http://scratch.proteomics.ics.uci.edu.

MITOMAP (http://www.MITOMAP.org), a database for the human mitochondrial genome, has grown rapidly in data content over the past several years as interest in the role of mitochondrial DNA (mtDNA) variation in human origins, forensics, degenerative diseases, cancer and aging has increased dramatically. To accommodate this information explosion, MITOMAP has implemented a new relational database and an improved search engine, and all programs have been rewritten. System administrative changes have been made to improve security and efficiency, and to make MITOMAP compatible with a new automatic mtDNA sequence analyzer known as Mitomaster.

Circadian rhythms are intimately linked to cellular metabolism. Specifically, the NAD(+)-dependent deacetylase SIRT1, the founding member of the sirtuin family, contributes to clock function. Whereas SIRT1 exhibits diversity in deacetylation targets and subcellular localization, SIRT6 is the only constitutively chromatin-associated sirtuin and is prominently present at transcriptionally active genomic loci. Comparison of the hepatic circadian transcriptomes reveals that SIRT6 and SIRT1 separately control transcriptional specificity and therefore define distinctly partitioned classes of circadian genes. SIRT6 interacts with CLOCK:BMAL1 and, differently from SIRT1, governs their chromatin recruitment to circadian gene promoters. Moreover, SIRT6 controls circadian chromatin recruitment of SREBP-1, resulting in the cyclic regulation of genes implicated in fatty acid and cholesterol metabolism. This mechanism parallels a phenotypic disruption in fatty acid metabolism in SIRT6 null mice as revealed by circadian metabolome analyses. Thus, genomic partitioning by two independent sirtuins contributes to differential control of circadian metabolism.

MotivationAccurately predicting protein side-chain conformations is an important subproblem of the broader protein structure prediction problem. Several methods exist for generating fairly accurate models for moderate-size proteins in seconds or less. However, a major limitation of these methods is their inability to model post-translational modifications (PTMs) and unnatural amino acids. In natural living systems, the chemical groups added following translation are often critical for the function of the protein. In engineered systems, unnatural amino acids are incorporated into proteins to explore structure-function relationships and create novel proteins.ResultsWe present a new version of SIDEpro to predict the side chains of proteins containing non-standard amino acids, including 15 of the most frequently observed PTMs in the Protein Data Bank and all types of phosphorylation. SIDEpro uses energy functions that are parameterized by neural networks trained from available data. For PTMs, the [Formula: see text] and [Formula: see text] accuracies are comparable with those obtained for the precursor amino acid, and so are the RMSD values for the atoms shared with the precursor amino acid. In addition, SIDEpro can accommodate any PTM or unnatural amino acid, thus providing a flexible prediction system for high-throughput modeling of proteins beyond the standard amino acids.Availability and implementationSIDEpro programs and Web server, rotamer libraries and data are available through the SCRATCH suite of protein structure predictors at http://scratch.proteomics.ics.uci.edu/

Dropout is a recently introduced algorithm for training neural network by randomly dropping units during training to prevent their co-adaptation. A mathematical analysis of some of the static and dynamic properties of dropout is provided using Bernoulli gating variables, general enough to accommodate dropout on units or connections, and with variable rates. The framework allows a complete analysis of the ensemble averaging properties of dropout in linear networks, which is useful to understand the non-linear case. The ensemble averaging properties of dropout in non-linear logistic networks result from three fundamental equations: (1) the approximation of the expectations of logistic functions by normalized geometric means, for which bounds and estimates are derived; (2) the algebraic equality between normalized geometric means of logistic functions with the logistic of the means, which mathematically characterizes logistic functions; and (3) the linearity of the means with respect to sums, as well as products of independent variables. The results are also extended to other classes of transfer functions, including rectified linear functions. Approximation errors tend to cancel each other and do not accumulate. Dropout can also be connected to stochastic neurons and used to predict firing rates, and to backpropagation by viewing the backward propagation as ensemble averaging in a dropout linear network. Moreover, the convergence properties of dropout can be understood in terms of stochastic gradient descent. Finally, for the regularization properties of dropout, the expectation of the dropout gradient is the gradient of the corresponding approximation ensemble, regularized by an adaptive weight decay term with a propensity for self-consistent variance minimization and sparse representations.