Linking the epigenome to the genome: correlation of different features to DNA methylation of CpG islands.

DNA methylation of CpG islands plays a crucial role in the regulation of gene expression. More than half of all human promoters contain CpG islands with a tissue-specific methylation pattern in differentiated cells. Still today, the whole process of how DNA methyltransferases determine which region should be methylated is not completely revealed. There are many hypotheses of which genomic features are correlated to the epigenome that have not yet been evaluated. Furthermore, many explorative approaches of measuring DNA methylation are limited to a subset of the genome and thus, cannot be employed, e.g., for genome-wide biomarker prediction methods. In this study, we evaluated the correlation of genetic, epigenetic and hypothesis-driven features to DNA methylation of CpG islands. To this end, various binary classifiers were trained and evaluated by cross-validation on a dataset comprising DNA methylation data for 190 CpG islands in HEPG2, HEK293, fibroblasts and leukocytes. We achieved an accuracy of up to 91% with an MCC of 0.8 using ten-fold cross-validation and ten repetitions. With these models, we extended the existing dataset to the whole genome and thus, predicted the methylation landscape for the given cell types. The method used for these predictions is also validated on another external whole-genome dataset. Our results reveal features correlated to DNA methylation and confirm or disprove various hypotheses of DNA methylation related features. This study confirms correlations between DNA methylation and histone modifications, DNA structure, DNA sequence, genomic attributes and CpG island properties. Furthermore, the method has been validated on a genome-wide dataset from the ENCODE consortium. The developed software, as well as the predicted datasets and a web-service to compare methylation states of CpG islands are available at http://www.cogsys.cs.uni-tuebingen.de/software/dna-methylation/.

4D Flexible Atom-Pairs: An efficient probabilistic conformational space comparison for ligand-based virtual screening.

BACKGROUND: The performance of 3D-based virtual screening similarity functions is affected by the applied conformations of compounds. Therefore, the results of 3D approaches are often less robust than 2D approaches. The application of 3D methods on multiple conformer data sets normally reduces this weakness, but entails a significant computational overhead. Therefore, we developed a special conformational space encoding by means of Gaussian mixture models and a similarity function that operates on these models. The application of a model-based encoding allows an efficient comparison of the conformational space of compounds. RESULTS: Comparisons of our 4D flexible atom-pair approach with over 15 state-of-the-art 2D- and 3D-based virtual screening similarity functions on the 40 data sets of the Directory of Useful Decoys show a robust performance of our approach. Even 3D-based approaches that operate on multiple conformers yield inferior results. The 4D flexible atom-pair method achieves an averaged AUC value of 0.78 on the filtered Directory of Useful Decoys data sets. The best 2D- and 3D-based approaches of this study yield an AUC value of 0.74 and 0.72, respectively. As a result, the 4D flexible atom-pair approach achieves an average rank of 1.25 with respect to 15 other state-of-the-art similarity functions and four different evaluation metrics. CONCLUSIONS: Our 4D method yields a robust performance on 40 pharmaceutically relevant targets. The conformational space encoding enables an efficient comparison of the conformational space. Therefore, the weakness of the 3D-based approaches on single conformations is circumvented. With over 100,000 similarity calculations on a single desktop CPU, the utilization of the 4D flexible atom-pair in real-world applications is feasible.

Interpreting linear support vector machine models with heat map molecule coloring.

BACKGROUND: Model-based virtual screening plays an important role in the early drug discovery stage. The outcomes of high-throughput screenings are a valuable source for machine learning algorithms to infer such models. Besides a strong performance, the interpretability of a machine learning model is a desired property to guide the optimization of a compound in later drug discovery stages. Linear support vector machines showed to have a convincing performance on large-scale data sets. The goal of this study is to present a heat map molecule coloring technique to interpret linear support vector machine models. Based on the weights of a linear model, the visualization approach colors each atom and bond of a compound according to its importance for activity. RESULTS: We evaluated our approach on a toxicity data set, a chromosome aberration data set, and the maximum unbiased validation data sets. The experiments show that our method sensibly visualizes structure-property and structure-activity relationships of a linear support vector machine model. The coloring of ligands in the binding pocket of several crystal structures of a maximum unbiased validation data set target indicates that our approach assists to determine the correct ligand orientation in the binding pocket. Additionally, the heat map coloring enables the identification of substructures important for the binding of an inhibitor. CONCLUSIONS: In combination with heat map coloring, linear support vector machine models can help to guide the modification of a compound in later stages of drug discovery. Particularly substructures identified as important by our method might be a starting point for optimization of a lead compound. The heat map coloring should be considered as complementary to structure based modeling approaches. As such, it helps to get a better understanding of the binding mode of an inhibitor.

Modeling and benchmark data set for the inhibition of c-Jun N-terminal kinase-3.

The goal of this paper is to present and describe a novel 2D- and 3D-QSAR (quantitative structure-activity relationship) binary classification data set for the inhibition of c-Jun N-terminal kinase-3 with previously unpublished activities for a diverse set of compounds. JNK3 is an important pharmaceutical target because it is involved in many neurological disorders. Accordingly, the development of JNK3 inhibitors has gained increasing interest. 2D and 3D versions of the data set were used, consisting of 313 (70 actives) and 249 (60 actives) compounds, respectively. All compounds, for which activity was only determined for the racemate, were removed from the 3D data set. We investigated the diversity of the data sets by an agglomerative clustering with feature trees and show that the data set contains several different scaffolds. Furthermore, we show that the benchmarks can be tackled with standard supervised learning algorithms with a convincing performance. For the 2D problem, a random decision forest classifier achieves a Matthew's correlation coefficient of 0.744, the 3D problem could be modeled with a Matthew's correlation coefficient of 0.524 with 3D pharmacophores and a support vector machine. The performance of both data sets was evaluated within a nested 10-fold cross-validation. We therefore suggest that the data set is a reasonable basis for generating QSAR models for JNK3 because of its diverse composition and the performance of the classifiers presented in this study.

jCompoundMapper: An open source Java library and command-line tool for chemical fingerprints.

BACKGROUND: The decomposition of a chemical graph is a convenient approach to encode information of the corresponding organic compound. While several commercial toolkits exist to encode molecules as so-called fingerprints, only a few open source implementations are available. The aim of this work is to introduce a library for exactly defined molecular decompositions, with a strong focus on the application of these features in machine learning and data mining. It provides several options such as search depth, distance cut-offs, atom- and pharmacophore typing. Furthermore, it provides the functionality to combine, to compare, or to export the fingerprints into several formats. RESULTS: We provide a Java 1.6 library for the decomposition of chemical graphs based on the open source Chemistry Development Kit toolkit. We reimplemented popular fingerprinting algorithms such as depth-first search fingerprints, extended connectivity fingerprints, autocorrelation fingerprints (e.g. CATS2D), radial fingerprints (e.g. Molprint2D), geometrical Molprint, atom pairs, and pharmacophore fingerprints. We also implemented custom fingerprints such as the all-shortest path fingerprint that only includes the subset of shortest paths from the full set of paths of the depth-first search fingerprint. As an application of jCompoundMapper, we provide a command-line executable binary. We measured the conversion speed and number of features for each encoding and described the composition of the features in detail. The quality of the encodings was tested using the default parametrizations in combination with a support vector machine on the Sutherland QSAR data sets. Additionally, we benchmarked the fingerprint encodings on the large-scale Ames toxicity benchmark using a large-scale linear support vector machine. The results were promising and could often compete with literature results. On the large Ames benchmark, for example, we obtained an AUC ROC performance of 0.87 with a reimplementation of the extended connectivity fingerprint. This result is comparable to the performance achieved by a non-linear support vector machine using state-of-the-art descriptors. On the Sutherland QSAR data set, the best fingerprint encodings showed a comparable or better performance on 5 of the 8 benchmarks when compared against the results of the best descriptors published in the paper of Sutherland et al. CONCLUSIONS: jCompoundMapper is a library for chemical graph fingerprints with several tweaking possibilities and exporting options for open source data mining toolkits. The quality of the data mining results, the conversion speed, the LPGL software license, the command-line interface, and the exporters should be useful for many applications in cheminformatics like benchmarks against literature methods, comparison of data mining algorithms, similarity searching, and similarity-based data mining.

Large-scale learning of structure-activity relationships using a linear support vector machine and problem-specific metrics.

The goal of this study was to adapt a recently proposed linear large-scale support vector machine to large-scale binary cheminformatics classification problems and to assess its performance on various benchmarks using virtual screening performance measures. We extended the large-scale linear support vector machine library LIBLINEAR with state-of-the-art virtual high-throughput screening metrics to train classifiers on whole large and unbalanced data sets. The formulation of this linear support machine has an excellent performance if applied to high-dimensional sparse feature vectors. An additional advantage is the average linear complexity in the number of non-zero features of a prediction. Nevertheless, the approach assumes that a problem is linearly separable. Therefore, we conducted an extensive benchmarking to evaluate the performance on large-scale problems up to a size of 175000 samples. To examine the virtual screening performance, we determined the chemotype clusters using Feature Trees and integrated this information to compute weighted AUC-based performance measures and a leave-cluster-out cross-validation. We also considered the BEDROC score, a metric that was suggested to tackle the early enrichment problem. The performance on each problem was evaluated by a nested cross-validation and a nested leave-cluster-out cross-validation. We compared LIBLINEAR against a Naïve Bayes classifier, a random decision forest classifier, and a maximum similarity ranking approach. These reference approaches were outperformed in a direct comparison by LIBLINEAR. A comparison to literature results showed that the LIBLINEAR performance is competitive but without achieving results as good as the top-ranked nonlinear machines on these benchmarks. However, considering the overall convincing performance and computation time of the large-scale support vector machine, the approach provides an excellent alternative to established large-scale classification approaches.

Estimation of the applicability domain of kernel-based machine learning models for virtual screening.

BACKGROUND: The virtual screening of large compound databases is an important application of structural-activity relationship models. Due to the high structural diversity of these data sets, it is impossible for machine learning based QSAR models, which rely on a specific training set, to give reliable results for all compounds. Thus, it is important to consider the subset of the chemical space in which the model is applicable. The approaches to this problem that have been published so far mostly use vectorial descriptor representations to define this domain of applicability of the model. Unfortunately, these cannot be extended easily to structured kernel-based machine learning models. For this reason, we propose three approaches to estimate the domain of applicability of a kernel-based QSAR model. RESULTS: We evaluated three kernel-based applicability domain estimations using three different structured kernels on three virtual screening tasks. Each experiment consisted of the training of a kernel-based QSAR model using support vector regression and the ranking of a disjoint screening data set according to the predicted activity. For each prediction, the applicability of the model for the respective compound is quantitatively described using a score obtained by an applicability domain formulation. The suitability of the applicability domain estimation is evaluated by comparing the model performance on the subsets of the screening data sets obtained by different thresholds for the applicability scores. This comparison indicates that it is possible to separate the part of the chemspace, in which the model gives reliable predictions, from the part consisting of structures too dissimilar to the training set to apply the model successfully. A closer inspection reveals that the virtual screening performance of the model is considerably improved if half of the molecules, those with the lowest applicability scores, are omitted from the screening. CONCLUSION: The proposed applicability domain formulations for kernel-based QSAR models can successfully identify compounds for which no reliable predictions can be expected from the model. The resulting reduction of the search space and the elimination of some of the active compounds should not be considered as a drawback, because the results indicate that, in most cases, these omitted ligands would not be found by the model anyway.

Probabilistic Modeling of Conformational Space for 3D Machine Learning Approaches.

We present a new probabilistic encoding of the conformational space of a molecule that allows for the integration into common similarity calculations. The method uses distance profiles of flexible atom-pairs and computes generative models that describe the distance distribution in the conformational space. The generative models permit the use of probabilistic kernel functions and, therefore, our approach can be used to extend existing 3D molecular kernel functions, as applied in support vector machines, to build QSAR models. The resulting kernels are valid 4D kernel functions and reduce the dependency of the model quality on suitable conformations of the molecules. We showed in several experiments the robust performance of the 4D kernel function, which was extended by our approach, in comparison to the original 3D-based kernel function. The new method compares the conformational space of two molecules within one kernel evaluation. Hence, the number of kernel evaluations is significantly reduced in comparison to common kernel-based conformational space averaging techniques. Additionally, the performance gain of the extended model correlates with the flexibility of the data set and enables an a priori estimation of the model improvement.

Atomic local neighborhood flexibility incorporation into a structured similarity measure for QSAR.

In this work, we introduce a new method to regard the geometry in a structural similarity measure by approximating the conformational space of a molecule. Our idea is to break down the molecular conformation into the local conformations of neighbor atoms with respect to core atoms. This local geometry can be implicitly accessed by the trajectories of the neighboring atoms, which are emerge by rotatable bonds. In our approach, the physicochemical atomic similarity, which can be used in structured similarity measures, is augmented by a local flexibility similarity, which gives a rough estimate of the similarity of the local conformational space. We incorporated this new type of encoding the flexibility into the optimal assignment molecular similarity approach, which can be used as a pseudokernel in support vector machines. The impact of the local flexibility was evaluated on several published QSAR data sets. This lead to an improvement of the model quality on 9 out of 10 data sets compared to the unmodified optimal assignment kernel.

Optimal assignment methods for ligand-based virtual screening.

BACKGROUND: Ligand-based virtual screening experiments are an important task in the early drug discovery stage. An ambitious aim in each experiment is to disclose active structures based on new scaffolds. To perform these "scaffold-hoppings" for individual problems and targets, a plethora of different similarity methods based on diverse techniques were published in the last years. The optimal assignment approach on molecular graphs, a successful method in the field of quantitative structure-activity relationships, has not been tested as a ligand-based virtual screening method so far. RESULTS: We evaluated two already published and two new optimal assignment methods on various data sets. To emphasize the "scaffold-hopping" ability, we used the information of chemotype clustering analyses in our evaluation metrics. Comparisons with literature results show an improved early recognition performance and comparable results over the complete data set. A new method based on two different assignment steps shows an increased "scaffold-hopping" behavior together with a good early recognition performance. CONCLUSION: The presented methods show a good combination of chemotype discovery and enrichment of active structures. Additionally, the optimal assignment on molecular graphs has the advantage to investigate and interpret the mappings, allowing precise modifications of internal parameters of the similarity measure for specific targets. All methods have low computation times which make them applicable to screen large data sets.