Discerning intersecting fusion-activation pathways in the Nipah virus using machine learning.

The fusion of Nipah with host cells is facilitated by two of their glycoproteins, the G and the F proteins. The binding of cellular ephrins to the G head domain causes the G stalk domain to interact differently with F, which activates F to mediate virus-host fusion. To gain insight into how the ephrin-binding signal transduces from the head to the stalk domain of G, we examine quantitatively the differences between the conformational ensembles of the G head domain in its ephrin-bound and unbound states. We consider the human ephrins B2 and B3, and a double mutant of B2, all of which trigger fusion. The ensembles are generated using molecular dynamics, and the differences between them are quantified using a new machine learning method. We find that the portion of the G head domain whose conformational density is altered equivalently by the three ephrins is large, and comprises ∼25% of the residues in the G head domain. This subspace also includes the residues that are known to be important to F activation, which suggests that it contains at least one common signaling pathway. The spatial distribution of the residues constituting this subspace supports the model of signal transduction in which the signal transduces via the G head dimer interface. This study also adds to the growing list of examples where signaling does not depend solely on backbone deviations. In general, this study provides an approach to filter out conserved patterns in protein dynamics.

Duplicate gene enrichment and expression pattern diversification in multicellularity.

The enrichment of duplicate genes, and therefore paralogs (proteins coded by duplicate genes), in multicellular versus unicellular organisms enhances genomic functional innovation. This study quantitatively examined relationships among paralog enrichment, expression pattern diversification and multicellularity, aiming to better understand genomic basis of multicellularity. Paralog abundance in specific cells was compared with those in unicellular proteomes and the whole proteomes of multicellular organisms. The budding yeast, Saccharomyces cerevisiae and the nematode, Caenorhabditis elegans, for which the gene sets expressed in specific cells are available, were used as uni and multicellular models, respectively. Paralog count (K) distributions [P((k))] follow a power-law relationship [Formula in text] in the whole proteomes of both species and in specific C. elegans cells. The value of the constant α can be used as a gauge of paralog abundance; the higher the value, the lower the paralog abundance. The α-value is indeed lower in the whole proteome of C. elegans (1.74) than in S. cerevisiae (2.34), quantifying the enrichment of paralogs in multicellular species. We also found that the power-law relationship applies to the proteomes of specific C. elegans cells. Strikingly, values of α in specific cells are higher and comparable to that in S. cerevisiae. Thus, paralog abundance in specific cells is lower and comparable to that in unicellular species. Furthermore, how much the expression level of a gene fluctuates across different C. elegans cells correlates positively with its paralog count, which is further confirmed by human gene-expression patterns across different tissues. Taken together, these results quantitatively and mechanistically establish enrichment of paralogs with diversifying expression patterns as genomic and evolutionary basis of multicellularity.

Use of artificial neural networks to determine cognitive impairment and therapeutic effectiveness in Alzheimer's transgenic mice.

Behavioral testing of transgenic mouse models of Alzheimer's disease (AD) is the functional endpoint for determining the effectiveness of therapeutic interventions and elucidating AD pathogenesis. Utilizing these mouse models, there have been remarkably few attempts to analyze multiple behavioral measures/tasks with higher-level computation techniques, either to distinguish performance between transgenic groups or to reveal any "overall" cognitive benefit of a given therapeutic. The present study compared the classificatory accuracy of artificial neural networks (ANNs) versus more traditional discriminant function analysis (DFA) using multiple behavioral measures/tasks from two AD transgenic mouse investigations. These investigations were to determine if AD transgenic mice could be cognitively-protected by either long-term caffeine administration (CA) or by a cognitively-stimulating environment (SE). Both the entire set of behavioral measures and a subset of 8 cognitive-based measures were analyzed. Both classifiers revealed a beneficial "overall" effect of CA and SE to protect AD transgenic mice across multiple cognitive measures/tasks. However, for both CA and SE datasets, the ANN was superior to DFA for discerning transgenicity (non-transgenic vs. transgenic-controls) across multiple behavioral measures. These results indicate that ANNs have an excellent capacity to discriminate cognitive impairment in AD transgenic mice and thus designate ANNs as a novel, sensitive method for cognitive assessment in Alzheimer's research.

Cognitive impairment in PDAPP mice depends on ApoE and ACT-catalyzed amyloid formation.

Biochemical and genetic studies indicate that the inflammatory proteins, apolipoprotein E (ApoE) and alpha(1)-antichymotrypsin (ACT) are important in the pathogenesis of Alzheimer's disease (AD). Using several lines of multiply transgenic/knockout mice we show here that murine ApoE and human ACT separately and synergistically facilitate both diffuse A beta immunoreactive and fibrillar amyloid deposition and thus also promote cognitive impairment in aged PDAPP(V717F) mice. The degree of cognitive impairment is highly correlated with the ApoE- and ACT-dependent hippocampal amyloid burden, with PDAPP mice lacking ApoE and ACT having little amyloid and little learning disability. A analysis of young mice before the onset of amyloid formation shows that steady-state levels of monomeric A beta peptide are unchanged by ApoE or ACT. These data suggest that the process or product of amyloid formation is more critical than monomeric A beta for the neurological decline in AD, and that the risk factors ApoE and ACT participate primarily in disease processes downstream of APP processing.

Use of multimetric statistical analysis to characterize and discriminate between the performance of four Alzheimer's transgenic mouse lines differing in Abeta deposition.

Behavioral assessment of genetically-manipulated mouse lines for Alzheimer's disease has become an important index for determining the efficacy of therapeutic interventions and examining disease pathogenesis. However, the potential for higher level statistical analyses to assist in these goals remains largely unexplored. The present study thus involved multimetric statistical analyses of behavioral and beta-amyloid (Abeta) deposition measures from four PDAPP-derived transgenic mouse lines that differ in extent of Abeta deposition. For all four lines, multiple behavioral measures obtained from a comprehensive task battery administered at 15-16 months of age were collectively examined by correlation, factor, and discriminant function analyses. In addition, both compact and total beta-amyloid (Abeta) histologic measures were determined from the same animals. Widespread intra- and inter-task correlations were evident, with impairment in all three water tasks (Morris maze, platform recognition, and radial arm water maze) correlating extensively with Abeta deposition in hippocampus and cerebral cortex. By elucidating the underlying relationships among measures, factor analysis revealed a single primary factor (Factor 1) that loaded most cognitive measures, particularly those for working memory and recognition. Abeta deposition measures loaded exclusively on this primary factor. In individual animals, only factor scores derived from this primary factor were correlated with Abeta deposition. Both of these findings again underscore the association between cognitive impairment and Abeta deposition. Finally, discriminant function analysis (step-wise forward method) was able to distinguish between all four AD transgenic lines based on behavioral performance alone, as well as when Abeta deposition measures were included. Our results demonstrate the utility of higher level, multimetric analysis of behavioral measures from AD transgenic mice. Analyses such as these will be very beneficial for the functional evaluation of therapeutic interventions against AD and for finding behavioral measures that can serve as predictors of pathology.

Multi-metric behavioral comparison of APPsw and P301L models for Alzheimer's disease: linkage of poorer cognitive performance to tau pathology in forebrain.

APPsw transgenic mice bearing the "Swedish" amyloid precursor protein (APP) mutation and JNPL3 transgenic mice bearing the P301L (Tau) mutation were compared to control non-transgenic (NT) mice in an extensive behavioral test battery administered between 5 and 8.5 months of age. APP mice were impaired in a variety of cognitive-based tasks prior to overt Abeta plaque development, making involvement of mutant APP overexpression and/or oligomeric Abeta assemblies most likely. Although Tau mice, as a group, were not impaired in any single behavioral measure, a collective assessment of behavioral measures through discriminant function analysis showed that Tau mice were impaired in overall behavioral (cognitive) performance. Moreover, correlation analyses involving Tau mice alone revealed linkage between poorer cognitive performance in all three water maze tasks and the number of neurofibrillary tangle (NFT)-containing neurons in neocortex and hippocampus. These findings indicate that: (1) APP mice show early and extensive cognitive impairment before evident Abeta deposition, and (2) the process or product of NFT formation in Tau mice is sufficient to deleteriously impact cognitive performance.

Quantifying Changes in Intrinsic Molecular Motion Using Support Vector Machines.

The ensemble of three-dimensional (3-D) configurations exhibited by a molecule, that is, its intrinsic motion, can be altered by several environmental factors, and also by the binding of other molecules. Quantification of such induced changes in intrinsic motion is important because it provides a basis for relating thermodynamic changes to changes in molecular motion. This task is, however, challenging because it requires comparing two high-dimensional data sets. Traditionally, when analyzing molecular simulations, this problem is circumvented by first reducing the dimensions of the two ensembles separately, and then comparing summary statistics from the two ensembles against each other. However, since dimensionality reduction is carried out prior to ensemble comparison, such strategies are susceptible to artifactual biases from information loss. Here, we introduce a method based on support vector machines that yields a normalized quantitative estimate for the difference between two ensembles after comparing them directly against one another. While this method can be applied to any molecular system, including nonbiological molecules and crystals, here, we show how it can be applied to identify the specific regions of a paramyxovirus G protein that are affected by the binding of its preferred human receptor, Ephrin B2. This protein-protein interaction initiates the fusion of the virus with the host cell. Specifically, for every residue in the G protein, we obtain separately a quantitative difference between the ensemble of configurations they sample in the presence and in the absence of Ephrin B2. These ensembles were generated using molecular dynamics simulations. Rank-ordering and then mapping the residues that undergo the greatest change in motion onto the 3-D structure of the G protein reveals that they are clustered primarily on a single contiguous facet of the protein and include the set that is known experimentally to play a vital role in regulating viral fusion.