Analysis of the Internal Representations Developed by Neural Networks for Structures Applied to Quantitative Structure−Activity Relationship Studies of Benzodiazepines

Micheli, Alessio; Sperduti, Alessandro; Starita, Antonina; Bianucci, Anna Maria Paola

doi:10.1021/ci9903399

Cited by 67 publications

(61 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, running the experiment over the same training and testing split as in Micheli et al, 37,39 we obtain a correlation coefficient of 0.98, which is not significantly different from the 0.999 value reported by these authors. This kernel is more likely to capture the global structure of the molecules and their chemical groups.…”

Section: Small Data Setscontrasting

confidence: 41%

See 1 more Smart Citation

One‐ to Four‐Dimensional Kernels for Virtual Screening and the Prediction of Physical, Chemical, and Biological Properties.

Azencott¹,

Ksikes²,

Swamidass³

et al. 2007

ChemInform

View full text Add to dashboard Cite

Many chemoinformatics applications, including high-throughput virtual screening, benefit from being able to rapidly predict the physical, chemical, and biological properties of small molecules to screen large repositories and identify suitable candidates. When training sets are available, machine learning methods provide an effective alternative to ab initio methods for these predictions. Here, we leverage rich molecular representations including 1D SMILES strings, 2D graphs of bonds, and 3D coordinates to derive efficient machine learning kernels to address regression problems. We further expand the library of available spectral kernels for small molecules developed for classification problems to include 2.5D surface and 3D kernels using Delaunay tetrahedrization and other techniques from computational geometry, 3D pharmacophore kernels, and 3.5D or 4D kernels capable of taking into account multiple molecular configurations, such as conformers. The kernels are comprehensively tested using cross-validation and redundancy-reduction methods on regression problems using several available data sets to predict boiling points, melting points, aqueous solubility, octanol/water partition coefficients, and biological activity with state-of-the art results. When sufficient training data are available, 2D spectral kernels in general tend to yield the best and most robust results, better than state-of-the art. On data sets containing thousands of molecules, the kernels achieve a squared correlation coefficient of 0.91 for aqueous solubility prediction and 0.94 for octanol/water partition coefficient prediction. Averaging over conformations improves the performance of kernels based on the three-dimensional structure of molecules, especially on challenging data sets. Kernel predictors for aqueous solubility (kSOL), LogP (kLOGP), and melting point (kMELT) are available over the Web through: http:// cdb.ics.uci.edu.

show abstract

Section: Small Data Setscontrasting

confidence: 41%

“…Although it is rather small, it is characterized by good molecular diversity, which is significant for QSAR analysis. The best performance on this data set is reported in Micheli et al 37 with a correlation coefficient of 0.999 obtained by a cascade correlation neural network.…”

Section: Datamentioning

confidence: 76%

One‐ to Four‐Dimensional Kernels for Virtual Screening and the Prediction of Physical, Chemical, and Biological Properties.

Azencott¹,

Ksikes²,

Swamidass³

et al. 2007

ChemInform

View full text Add to dashboard Cite

show abstract

“…Object localization (Bianchini et al 2005), image classification (Francesconi et al 1998), natural language processing (Krahmer et al 2003), bioinformatics (Baldi and Pollastri 2004), QSAR (Micheli et al 2001), web page scoring, social network analysis (Newman 2001) and relational learning are examples of application domains where the information of interest is encoded into a set of basic entities and relationships between them. In all these domains, the data is naturally represented by sequences, trees, and, more generally, directed or undirected graphs.…”

Section: Introductionmentioning

confidence: 99%

“…Supervised connectionist models for graph processing have been used in several application domains, including protein structure prediction (Baldi and Pollastri 2004;Money and Pollastri 2009), QSAR (Micheli et al 2001), theorem proving (Goller 1997), image classification and object localization in images (Bianchini et al 2003;Di Massa et al 2006), language recognition (Rodriguez 2001;Sturt et al 2003), logo recognition (Francesconi et al 1998) and web page ranking . Applications for unsupervised methods include XML clustering and classification (Yong et al 2006;Hagenbuchner et al 2006), image classification (Wang et al 2002) and web document clustering (Bloehdorn and Blohm 2006).…”

mentioning

confidence: 99%

Neural networks for relational learning: an experimental comparison

et al. 2010

View full text Add to dashboard Cite

In the last decade, connectionist models have been proposed that can process structured information directly. These methods, which are based on the use of graphs for the representation of the data and the relationships within the data, are particularly suitable for handling relational learning tasks. In this paper, two recently proposed architectures of this kind, i.e. Graph Neural Networks (GNNs) and Relational Neural Networks (RelNNs), are compared and discussed, along with their corresponding learning schemes. The goal is to evaluate the performance of these methods on benchmarks that are commonly used by the relational learning community. Moreover, we also aim at reporting differences in the behavior of the two models, in order to gain insights on possible extensions of the approaches. Since RelNNs have been developed with the specific task of learning aggregate functions in mind, some experiments are run considering that particular task. In addition, we carry out more general experiments on the mutagenesis and the biodegradability datasets, on which several other relational learners have been evaluated. The experimental results are promising and suggest that RelNNs and GNNs can be a viable approach for learning on relational data.

show abstract

“…To extract meaning, patterns, and regularities from these data requires computational methods that can not only handle structured data but also leverage structural information. Two classes of machine learning methods that have been applied to structured data are probabilistic graphical models [14,10] such as Bayesian networks, and recursive neural networks [1,9,16,11,8,12,2]. The purpose of this article is to analyze the mathematical relationship between these two approaches and, in particular, to show how a recursive neural network can be viewed as a limit of, or a fast approximation to, a sequences of Bayesian networks.…”

Section: Introductionmentioning

confidence: 99%

On the relationship between deterministic and probabilistic directed Graphical models: From Bayesian networks to recursive neural networks

Baldi¹,

Rosen-Zvi²

2005

Neural Networks

View full text Add to dashboard Cite

Machine learning methods that can handle variable-size structured data such as sequences and graphs include Bayesian networks (BNs) and Recursive Neural Networks (RNNs). In both classes of models, the data is modeled using a set of observed and hidden variables associated with the nodes of a directed acyclic graph. In BNs, the conditional relationships between parent and child variables are probabilistic whereas in RNNs they are deterministic and parameterized by neural networks. Here we study the formal relationship between both classes of models and show that when the source nodes variables are observed, RNNs can be viewed as limits, both in distribution and probability, of BNs with local conditional distributions that have vanishing covariance matrices and converge to delta functions. Conditions for uniform convergence are also given together with an analysis of the behavior and exactness of Belief Propagation (BP) in "deterministic" BNs. Implications for the design of mixed architectures and the corresponding inference algorithms are briefly discussed.

show abstract

Analysis of the Internal Representations Developed by Neural Networks for Structures Applied to Quantitative Structure−Activity Relationship Studies of Benzodiazepines

Cited by 67 publications

References 33 publications

One‐ to Four‐Dimensional Kernels for Virtual Screening and the Prediction of Physical, Chemical, and Biological Properties.

One‐ to Four‐Dimensional Kernels for Virtual Screening and the Prediction of Physical, Chemical, and Biological Properties.

Neural networks for relational learning: an experimental comparison

On the relationship between deterministic and probabilistic directed Graphical models: From Bayesian networks to recursive neural networks

Contact Info

Product

Resources

About