Novel topological descriptors for analyzing biological networks

Network-based artificial intelligence approaches for advancing personalized psychiatry

Psychiatric disorders have a complex biological underpinning likely involving an interplay of genetic and environmental risk contributions. Substantial efforts are being made to use artificial intelligence approaches to integrate features within and across data types to increase our etiological understanding and advance personalized psychiatry. Network science offers a conceptual framework for exploring the often complex relationships across different levels of biological organization, from cellular mechanistic to brain-functional and phenotypic networks. Utilizing such network information effectively as part of artificial intelligence approaches is a promising route toward a more in-depth understanding of illness biology, the deciphering of patient heterogeneity, and the identification of signatures that may be sufficiently predictive to be clinically useful. Here, we present examples of how network information has been used as part of artificial intelligence within psychiatry and beyond and outline future perspectives on how personalized psychiatry approaches may profit from a closer integration of psychiatric research, artificial intelligence development, and network science.

A QSPR analysis of physical properties of antituberculosis drugs using neighbourhood degree-based topological indices and support vector regression

Topological indices are molecular descriptors used in QSPR modelling to predict the physicochemical properties of molecules. Topological indices are used in numerous applications in drug design. In this work, we compute the neighbourhood degree-based topological indices of 15 antituberculosis drugs, we studied the QSPR analysis of these drugs using support vector regression. The efficiency of support vector regression is determined by comparing it with the classical linear regression. Our QSPR model further shows the superiority of the SVR model as a better predictive model in QSPR analysis of the physical properties of antituberculosis drugs. The findings in this study are a further contribution to the field of chemical graph theory and drug design, providing a deeper understanding of neighbourhood degree-based topological indices and their predictive capabilities in QSPR model.

Predicting prediction: A systematic workflow to analyze factors affecting the classification performance in genomic biomarker discovery

High throughput technologies in genomics enable the analysis of small alterations in gene expression levels. Patterns of such deviations are an important starting point for the discovery and verification of new biomarker candidates. Identifying such patterns is a challenging task that requires sophisticated machine learning approaches. Currently, there are a variety of classification models, and a common approach is to compare the performance and select the best one for a given classification problem. Since the association between the features of a data set and the performance of a particular classification method is still not fully understood, the main contribution of this work is to provide a new methodology for predicting the prediction results of different classifiers in the field of biomarker discovery. We propose here a three-steps computational workflow that includes an analysis of the data set characteristics, the calculation of the classification accuracy and, finally, the prediction of the resulting classification error. The experiments were carried out on synthetic and microarray datasets. Using this method, we showed that the predictability strongly depends on the discriminatory ability of the features, e.g., sets of genes, in two or multi-class datasets. If a dataset has a certain discriminatory ability, this method enables prediction of the classification performance before applying a learning model. Thus, our results contribute to a better understanding of the relationship between dataset characteristics and the corresponding performance of a machine learning method, and suggest the optimal classification method for a given dataset based on its discriminatory ability.

Saagar-A New, Extensible Set of Molecular Substructures for QSAR/QSPR and Read-Across Predictions

Molecular structure-based predictive models provide a proven alternative to costly and inefficient animal testing. However, due to a lack of interpretability of predictive models built with abstract molecular descriptors they have earned the notoriety of being black boxes. Interpretable models require interpretable descriptors to provide chemistry-backed predictive reasoning and facilitate intelligent molecular design. We developed a novel set of extensible chemistry-aware substructures, Saagar, to support interpretable predictive models and read-across protocols. Performance of Saagar in chemical characterization and search for structurally similar actives for read-across applications was compared with four publicly available fingerprint sets (MACCS (166), PubChem (881), ECFP4 (1024), ToxPrint (729)) in three benchmark sets (MUV, ULS, and Tox21) spanning ∼145 000 compounds and 78 molecular targets at 1%, 2%, 5%, and 10% false discovery rates. In 18 of the 20 comparisons, interpretable Saagar features performed better than the publicly available, but less interpretable and fixed-bit length, fingerprints. Examples are provided to show the enhanced capability of Saagar in extracting compounds with higher scaffold similarity. Saagar features are interpretable and efficiently characterize diverse chemical collections, thus making them a better choice for building interpretable predictive in silico models and read-across protocols.

Properties of Entropy-Based Topological Measures of Fullerenes

A fullerene is a cubic three-connected graph whose faces are entirely composed of pentagons and hexagons. Entropy applied to graphs is one of the significant approaches to measuring the complexity of relational structures. Recently, the research on complex networks has received great attention, because many complex systems can be modelled as networks consisting of components as well as relations among these components. Information—theoretic measures have been used to analyze chemical structures possessing bond types and hetero-atoms. In the present article, we reviewed various entropy-based measures on fullerene graphs. In particular, we surveyed results on the topological information content of a graph, namely the orbit-entropy Ia(G), the symmetry index, a degree-based entropy measure Iλ(G), the eccentric-entropy Ifσ(G) and the Hosoya entropy H(G).

A Decomposition Method for Global Evaluation of Shannon Entropy and Local Estimations of Algorithmic Complexity

We investigate the properties of a Block Decomposition Method (BDM), which extends the power of a Coding Theorem Method (CTM) that approximates local estimations of algorithmic complexity based on Solomonoff–Levin’s theory of algorithmic probability providing a closer connection to algorithmic complexity than previous attempts based on statistical regularities such as popular lossless compression schemes. The strategy behind BDM is to find small computer programs that produce the components of a larger, decomposed object. The set of short computer programs can then be artfully arranged in sequence so as to produce the original object. We show that the method provides efficient estimations of algorithmic complexity but that it performs like Shannon entropy when it loses accuracy. We estimate errors and study the behaviour of BDM for different boundary conditions, all of which are compared and assessed in detail. The measure may be adapted for use with more multi-dimensional objects than strings, objects such as arrays and tensors. To test the measure we demonstrate the power of CTM on low algorithmic-randomness objects that are assigned maximal entropy (e.g., π) but whose numerical approximations are closer to the theoretical low algorithmic-randomness expectation. We also test the measure on larger objects including dual, isomorphic and cospectral graphs for which we know that algorithmic randomness is low. We also release implementations of the methods in most major programming languages—Wolfram Language (Mathematica), Matlab, R, Perl, Python, Pascal, C++, and Haskell—and an online algorithmic complexity calculator.

Descriptor selection based on variable stability for predicting inhibitor activity

Quantitative structure-activity relationship (QSAR) has been a technique to study the relationship between chemical structures and properties, and variable selection is an important problem for finding the informative variables and building reliable models. A variable selection method based on variable stability is proposed and used for selecting the informative descriptors in the QSAR model of inhibitors. In the method, a series of models are built by leave-one-out cross validation (LOOCV), and variable stability is defined as the ratio of the absolute mean value and standard deviation of the regression coefficients in the models for a descriptor. Therefore, the descriptors with larger stabilities are more informative to the model. To further enhance the difference among the descriptors, an exponential parameter is used to modify the standard deviation. The results show that 13 descriptors are selected as informative ones from 1217 descriptors for the QSAR model of inhibitors. An effective prediction model can be constructed by them.

Mutual information model for link prediction in heterogeneous complex networks

Recently, a number of meta-path based similarity indices like PathSim, HeteSim, and random walk have been proposed for link prediction in heterogeneous complex networks. However, these indices suffer from two major drawbacks. Firstly, they are primarily dependent on the connectivity degrees of node pairs without considering the further information provided by the given meta-path. Secondly, most of them are required to use a single and usually symmetric meta-path in advance. Hence, employing a set of different meta-paths is not straightforward. To tackle with these problems, we propose a mutual information model for link prediction in heterogeneous complex networks. The proposed model, called as Meta-path based Mutual Information Index (MMI), introduces meta-path based link entropy to estimate the link likelihood and could be carried on a set of available meta-paths. This estimation measures the amount of information through the paths instead of measuring the amount of connectivity between the node pairs. The experimental results on a Bibliography network show that the MMI obtains high prediction accuracy compared with other popular similarity indices.

Computational characterization of parallel dimeric and trimeric coiled-coils using effective amino acid indices

RFCoil is a novel predictor for parallel coiled-coil dimer and trimer.

The discrimination power of structural SuperIndices

In this paper, we evaluate the discrimination power of structural superindices. Superindices for graphs represent measures composed of other structural indices. In particular, we compare the discrimination power of the superindices with those of individual graph descriptors. In addition, we perform a statistical analysis to generalize our findings to large graphs.

RMol: a toolset for transforming SD/Molfile structure information into R objects

Background

The graph-theoretical analysis of molecular networks has a long tradition in chemoinformatics. As demonstrated frequently, a well designed format to encode chemical structures and structure-related information of organic compounds is the Molfile format. But when it comes to use modern programming languages for statistical data analysis in Bio- and Chemoinformatics, R as one of the most powerful free languages lacks tools to process Molfile data collections and import molecular network data into R.

Results

We design an R object which allows a lossless information mapping of structural information from Molfiles into R objects. This provides the basis to use the RMol object as an anchor for connecting Molfile data collections with R libraries for analyzing graphs. Associated with the RMol objects, a set of R functions completes the toolset to organize, describe and manipulate the converted data sets. Further, we bypass R-typical limits for manipulating large data sets by storing R objects in bz-compressed serialized files instead of employing RData files.

Conclusions

By design, RMol is a R toolset without dependencies to other libraries or programming languages. It is useful to integrate into pipelines for serialized batch analysis by using network data and, therefore, helps to process sdf-data sets in R efficiently. It is freely available under the BSD licence. The script source can be downloaded from http://sourceforge.net/p/rmol-toolset.

A network-based feature selection approach to identify metabolic signatures in disease

The identification and interpretation of metabolic biomarkers is a challenging task. In this context, network-based approaches have become increasingly a key technology in systems biology allowing to capture complex interactions in biological systems. In this work, we introduce a novel network-based method to identify highly predictive biomarker candidates for disease. First, we infer two different types of networks: (i) correlation networks, and (ii) a new type of network called ratio networks. Based on these networks, we introduce scores to prioritize features using topological descriptors of the vertices. To evaluate our method we use an example dataset where quantitative targeted MS/MS analysis was applied to a total of 52 blood samples from 22 persons with obesity (BMI >30) and 30 healthy controls. Using our network-based feature selection approach we identified highly discriminating metabolites for obesity (F-score >0.85, accuracy >85%), some of which could be verified by the literature.

Information indices with high discriminative power for graphs

In this paper, we evaluate the uniqueness of several information-theoretic measures for graphs based on so-called information functionals and compare the results with other information indices and non-information-theoretic measures such as the well-known Balaban J index. We show that, by employing an information functional based on degree-degree associations, the resulting information index outperforms the Balaban J index tremendously. These results have been obtained by using nearly 12 million exhaustively generated, non-isomorphic and unweighted graphs. Also, we obtain deeper insights on these and other topological descriptors when exploring their uniqueness by using exhaustively generated sets of alkane trees representing connected and acyclic graphs in which the degree of a vertex is at most four.

Profiling the human response to physical exercise: a computational strategy for the identification and kinetic analysis of metabolic biomarkers

BACKGROUND

In metabolomics, biomarker discovery is a highly data driven process and requires sophisticated computational methods for the search and prioritization of novel and unforeseen biomarkers in data, typically gathered in preclinical or clinical studies. In particular, the discovery of biomarker candidates from longitudinal cohort studies is crucial for kinetic analysis to better understand complex metabolic processes in the organism during physical activity.

FINDINGS

In this work we introduce a novel computational strategy that allows to identify and study kinetic changes of putative biomarkers using targeted MS/MS profiling data from time series cohort studies or other cross-over designs. We propose a prioritization model with the objective of classifying biomarker candidates according to their discriminatory ability and couple this discovery step with a novel network-based approach to visualize, review and interpret key metabolites and their dynamic interactions within the network. The application of our method on longitudinal stress test data revealed a panel of metabolic signatures, i.e., lactate, alanine, glycine and the short-chain fatty acids C2 and C3 in trained and physically fit persons during bicycle exercise.

CONCLUSIONS

We propose a new computational method for the discovery of new signatures in dynamic metabolic profiling data which revealed known and unexpected candidate biomarkers in physical activity. Many of them could be verified and confirmed by literature. Our computational approach is freely available as R package termed BiomarkeR under LGPL via CRAN http://cran.r-project.org/web/packages/BiomarkeR/.

A network-based approach to classify the three domains of life

Background

Identifying group-specific characteristics in metabolic networks can provide better insight into evolutionary developments. Here, we present an approach to classify the three domains of life using topological information about the underlying metabolic networks. These networks have been shown to share domain-independent structural similarities, which pose a special challenge for our endeavour. We quantify specific structural information by using topological network descriptors to classify this set of metabolic networks. Such measures quantify the structural complexity of the underlying networks. In this study, we use such measures to capture domain-specific structural features of the metabolic networks to classify the data set. So far, it has been a challenging undertaking to examine what kind of structural complexity such measures do detect. In this paper, we apply two groups of topological network descriptors to metabolic networks and evaluate their classification performance. Moreover, we combine the two groups to perform a feature selection to estimate the structural features with the highest classification ability in order to optimize the classification performance.

Results

By combining the two groups, we can identify seven topological network descriptors that show a group-specific characteristic by ANOVA. A multivariate analysis using feature selection and supervised machine learning leads to a reasonable classification performance with a weighted F-score of 83.7% and an accuracy of 83.9%. We further demonstrate that our approach outperforms alternative methods. Also, our results reveal that entropy-based descriptors show the highest classification ability for this set of networks.

Conclusions

Our results show that these particular topological network descriptors are able to capture domain-specific structural characteristics for classifying metabolic networks between the three domains of life.