Generative Topographic Mapping Research Articles

A Chemographic Audit of anti-Coronavirus Structure-activity Information from Public Databases (ChEMBL).

Discovery of drugs against newly emerged pathogenic agents like the SARS‐CoV‐2 coronavirus (CoV) must be based on previous research against related species. Scientists need to get acquainted with and develop a global oversight over so‐far tested molecules. Chemography (herein used Generative Topographic Mapping, in particular) places structures on a human‐readable 2D map (obtained by dimensionality reduction of the chemical space of molecular descriptors) and is thus well suited for such an audit. The goal is to map medicinal chemistry efforts so far targeted against CoVs. This includes comparing libraries tested against various virus species/genera, predicting their polypharmacological profiles and highlighting often encountered chemotypes. Maps are challenged to provide predictive activity landscapes against viral proteins. Definition of “anti‐CoV” map zones led to selection of therein residing 380 potential anti‐CoV agents, out of a vast pool of 800 M organic compounds.

Parallel Generative Topographic Mapping: An Efficient Approach for Big Data Handling.

Generative Topographic Mapping (GTM) can be efficiently used to visualize, analyze and model large chemical data. The GTM manifold needs to span the chemical space deemed relevant for a given problem. Therefore, the Frame set (FS) of compounds used for the manifold construction must well cover a given chemical space. Intuitively, the FS size must raise with the size and diversity of the target library. At the same time, the GTM training can be very slow or even becomes technically impossible at FS sizes of the order of 105 compounds – which is a very small number compared to today's commercially accessible compounds, and, especially, to the theoretically feasible molecules. In order to solve this problem, we propose a Parallel GTM algorithm based on the merging of “intermediate” manifolds constructed in parallel for different subsets of molecules. An ensemble of these subsets forms a FS for the “final” manifold. In order to assess the efficiency of the new algorithm, 80 GTMs were built on the FSs of different sizes ranging from 10 to 1.8 M compounds selected from the ChEMBL database. Each GTM was challenged to build classification models for up to 712 biological activities (depending on the FS size). With the novel parallel GTM procedure, we could thus cover the entire spectrum of possible FS sizes, whereas previous studies were forced to rely on the working hypothesis that FS sizes of few thousands of compounds are sufficient to describe the ChEMBL chemical space. In fact, this study formally proves this to be true: a FS containing only 5000 randomly picked compounds is sufficient to represent the entire ChEMBL collection (1.8 M molecules), in the sense that a further increase of FS compound numbers has no benefice impact on the predictive propensity of the above‐mentioned 712 activity classification models. Parallel GTM may, however, be required to generate maps based on very large FS, that might improve chemical space cartography of big commercial and virtual libraries, approaching billions of compounds

Parallel Generative Topographic Mapping: an Efficient Approach for Big Data Handling

Parallel Generative Topographic Mapping: an Efficient Approach for Big Data Handling

Application of the mol2vec Technology to Large-size Data Visualization and Analysis.

Generative Topographic Mapping (GTM) is a dimensionality reduction method, which is widely used for both data visualization and structure-activity modeling. Large dimensionality of the initial data space may require significant computational resources and slow down the GTM construction. Therefore, it may be meaningful to reduce the number of descriptors used for encoding molecular structures. The Principal Component Analysis (PCA), a standard preprocessing tool, suffers from the information loss upon the dimensionality reduction. As an alternative, we propose to use substructure vector embedding provided by the mol2vec technique. In addition to the data dimensionality reduction, this technology also accounts for proximity of substructures in molecular graphs. In this study, dimensionality of large descriptor spaces of ISIDA fragment descriptors or Morgan fingerprints were reduced using either the PCA or the mol2vec method. The latter significantly speeds up GTM training without compromising its predictive power in bioactivity classification tasks.

Visualization Framework for High-Dimensional Spatio-Temporal Hydrological Gridded Datasets using Machine-Learning Techniques

Numerical modelling increasingly generates massive, high-dimensional spatio-temporal datasets. Exploring such datasets relies on effective visualization. This study presents a generic workflow to (i) project high-dimensional spatio-temporal data on a two-dimensional (2D) plane accurately (ii) compare dimensionality reduction techniques (DRTs) in terms of resolution and computational efficiency (iii) represent 2D projection spatially using a 2D perceptually uniform background color map. Machine learning (ML) based DRTs for data visualization i.e., principal component analysis (PCA), generative topographic mapping (GTM), t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) are compared in terms of accuracy, resolution and computational efficiency to handle massive datasets. The accuracy of visualization is evaluated using a quality metric based on a co-ranking framework. The workflow is applied to an output of an Australian Water Resource Assessment (AWRA) model for Tasmania, Australia. The dataset consists of daily time series of nine components of the water balance at a 5 km grid cell resolution for the year 2017. The case study shows that PCA allows rapid visualization of global data structures, while t-SNE and UMAP allows more accurate representation of local trends. Furthermore, UMAP is computationally more efficient than t-SNE and least affected by the outliers compared to GTM.

Seismic attribute selection for machine-learning-based facies analysis

Interpreters face two main challenges in seismic facies analysis. The first challenge is to define, or “label,” the facies of interest. The second challenge is to select a suite of attributes that can differentiate a target facies from the background reflectivity. Our key objective is to determine which seismic attributes can best differentiate one class of chaotic seismic facies from another using modern machine-learning technology. Although simple 1D histograms provide a list of candidate attributes, they do not provide insight into the optimum number or combination of attributes. To address this limitation, we have conducted an exhaustive search whereby we represent the target and background training facies by high-dimensional Gaussian mixture models (GMMs) for each potential attribute combination. The first step is to choose candidate attributes that may be able to differentiate chaotic mass-transport deposits and salt diapirs from the more conformal, coherent background reflectors. The second step is to draw polygons around the target and background facies to provide the labeled data to be represented by GMMs. Maximizing the distance between all GMM facies pairs provides the optimum number and combination of attributes. We use generative topographic mapping to represent the high-dimensional attribute data by a lower dimensional 2D manifold. Each labeled facies provides a probability density function on the manifold that can be compared to the probability density function of each voxel, providing the likelihood that a given voxel is a member of each of the facies. Our first example maps chaotic seismic facies associated with the development of salt diapirs and minibasins. Our second example successfully delineates karst collapse underlying a shale resource play from north Texas.

Modelling of ready biodegradability based on combined public and industrial data sources

ABSTRACTThe European Registration, Evaluation, Authorization and Restriction of Chemical Substances Regulation, requires marketed chemicals to be evaluated for Ready Biodegradability (RB), considering in silico prediction as valid alternative to experimental testing. However, currently available models may not be relevant to predict compounds of industrial interest, due to accuracy and applicability domain restriction issues. In this work, we present a new and extended RB dataset (2830 compounds), issued by the merging of several public data sources. It was used to train classification models, which were externally validated and benchmarked against already-existing tools on a set of 316 compounds coming from the industrial context. New models showed good performances in terms of predictive power (Balance Accuracy (BA) = 0.74–0.79) and data coverage (83–91%). The Generative Topographic Mapping approach identified several chemotypes and structural motifs unique to the industrial dataset, highlighting for which chemical classes currently available models may have less reliable predictions. Finally, public and industrial data were merged into global dataset containing 3146 compounds. This is the biggest dataset reported in the literature so far, covering some chemotypes absent in the public data. Thus, predictive model developed on the Global dataset has larger applicability domain than the existing ones.

Generative topographic mapping in drug design

This is a review article of Generative Topographic Mapping (GTM) - a non-linear dimensionality reduction technique producing generative 2D maps of high-dimensional vector spaces - and its specific applications in Drug Design (chemical space cartography, compound library design and analysis, virtual screening, pharmacological profiling, de novo drug design, conformational space & docking interaction cartography, etc.) Written by chemoinformaticians for potential users among medicinal chemists and biologists, the article purposely avoids all underlying mathematics. First, the GTM concept is intuitively explained, based on the strong analogies with the rather popular Self-Organizing Maps (SOMs), which are well established library analysis tools. GTM is basically a fuzzy-logics-based generalization of SOMs. The second part of the review, some of published GTM applications in drug design are briefly revisited.

Consensus models to predict oral rat acute toxicity and validation on a dataset coming from the industrial context

ABSTRACTWe report predictive models of acute oral systemic toxicity representing a follow-up of our previous work in the framework of the NICEATM project. It includes the update of original models through the addition of new data and an external validation of the models using a dataset relevant for the chemical industry context. A regression model for LD50 and multi-class classification model for toxicity classes according to the Global Harmonized System categories were prepared. ISIDA descriptors were used to encode molecular structures. Machine learning algorithms included support vector machine (SVM), random forest (RF) and naïve Bayesian. Selected individual models were combined in consensus. The different datasets were compared using the generative topographic mapping approach. It appeared that the NICEATM datasets were lacking some relevant chemotypes for chemical industry. The new models trained on enlarged data sets have applicability domains (AD) sufficiently large to accommodate industrial compounds. The fraction of compounds inside the models’ AD increased from 58% (NICEATM model) to 94% (new model). The increase of training sets improved models’ prediction performance: RMSE values decreased from 0.56 to 0.47 and balanced accuracies increased from 0.69 to 0.71 for NICEATM and new models, respectively.

A machine learning approach based on generative topographic mapping for disruption prevention and avoidance at JET

The need for predictive capabilities greater than 95% with very limited false alarms are demanding requirements for reliable disruption prediction systems in tokamaks such as JET or, in the near future, ITER. The prediction of an upcoming disruption must be provided sufficiently in advance in order to apply effective disruption avoidance or mitigation actions to prevent the machine from being damaged.In this paper, following the typical machine learning workflow, a generative topographic mapping (GTM) of the operational space of JET has been built using a set of disrupted and regularly terminated discharges. In order to build the predictive model, a suitable set of dimensionless, machine-independent, physics-based features have been synthesized, which make use of 1D plasma profile information, rather than simple zero-D time series. The use of such predicting features, together with the power of the GTM in fitting the model to the data, obtains, in an unsupervised way, a 2D map of the multi-dimensional parameter space of JET, where it is possible to identify a boundary separating the region free from disruption from the disruption region. In addition to helping in operational boundaries studies, the GTM map can also be used for disruption prediction exploiting the potential of the developed GTM toolbox to monitor the discharge dynamics. Following the trajectory of a discharge on the map throughout the different regions, an alarm is triggered depending on the disruption risk of these regions. The proposed approach to predict disruptions has been evaluated on a training and an independent test set and achieves very good performance with only one tardive detection and a limited number of false detections. The warning times are suitable for avoidance purposes and, more important, the detections are consistent with physical causes and mechanisms that destabilize the plasma leading to disruptions.

Diversifying chemical libraries with generative topographic mapping.

Generative topographic mapping was used to investigate the possibility to diversify the in-house compounds collection of Boehringer Ingelheim (BI). For this purpose, a 2D map covering the relevant chemical space was trained, and the BI compound library was compared to the Aldrich-Market Select (AMS) database of more than 8M purchasable compounds. In order to discover new (sub)structures, the "AutoZoom" tool was developed and applied in order to analyze chemotypes of molecules residing in heavily populated zones of a map and to extract the corresponding maximum common substructures. A set of 401K new structures from the AMS database was retrieved and checked for drug-likeness and biological activity.

QSPR models for bioconcentration factor (BCF): are they able to predict data of industrial interest?

ABSTRACT The bioconcentration factor (BCF), a key parameter required by the REACH regulation, estimates the tendency for a xenobiotic to concentrate inside living organisms. In silico methods can be valid alternatives to costly data measurements. However, in the industrial context, these theoretical approaches may fail to predict BCF with reasonable accuracy. We analyzed whether models built on public data only have adequate performances when challenged to predict industrial compounds. A new set of 1129 compounds has been collected by merging publicly available datasets. Generative Topographic Mapping was employed to compare this chemical space with a set of new compounds issued from the industry. Some new chemotypes absent in the training set (such as siloxanes) have been detected. A new BCF model has been built using ISIDA (In SIlico design and Data Analysis) fragment descriptors, support vector regression and random forest machine-learning methods. It has been externally validated on: (i) collected data from the literature and (ii) industrial data. The latter also served as benchmark for the freely available tools VEGA, EPISuite, TEST, OPERA. New model performs (RMSE of 0.58 log BCF units) comparably to existing ones but benefits of an extended applicability, covering the industrial set chemical space (78% data coverage).

Generative Topographic Mapping of the Docking Conformational Space

Following previous efforts to render the Conformational Space (CS) of flexible compounds by Generative Topographic Mapping (GTM), this polyvalent mapping technique is here adapted to the docking problem. Contact fingerprints (CF) characterize ligands from the perspective of the binding site by monitoring protein atoms that are “touched” by those of the ligand. A “Contact” (CF) map was built by GTM-driven dimensionality reduction of the CF vector space. Alternatively, a “Hybrid” (Hy) map used a composite descriptor of CFs concatenated with ligand fragment descriptors. These maps indirectly represent the active site and integrate the binding information of multiple ligands. The concept is illustrated by a docking study into the ATP-binding site of CDK2, using the S4MPLE program to generate thousands of poses for each ligand. Both maps were challenged to (1) Discriminate native from non-native ligand poses, e.g., create RMSD-landscapes “colored” by the conformer ensemble of ligands of known binding modes in order to highlight “native” map zones (poses with RMSD to PDB structures < 2Å). Then, projection of poses of other ligands on such landscapes might serve to predict those falling in native zones as being well-docked. (2) Distinguish ligands–characterized by their ensemble of conformers–by their potency, e.g., testing the hypotheses whether zones privileged by potent binders are clearly separated from the ones preferred by decoys on the maps. Hybrid maps were better in both challenges and outperformed the classical energy and individual contact satisfaction scores in discriminating ligands by potency. Moreover, the intuitive visualization and analysis of docking CS may, as already mentioned, have several applications–from highlighting of key contacts to monitoring docking calculation convergence.

Prediction of the Glass-Transition Temperatures of Linear Homo/Heteropolymers and Cross-Linked Epoxy Resins

This work proposes a unified approach to predict glass-transition temperatures (Tg) of linear homo/heteropolymers and cross-linked epoxy resins by machine-learning approaches based on descriptors of reagents undergoing polymerization, represented in a formal way such as to encompass all three scenarios: linear homo- and heteropolymers plus network heteropolymers. The “formal” representation of reagents is a problem-specific, herein designed standardization protocol of compounds, unlike typical structure curation rules in chemoinformatics. For example, heteropolymers are represented by the two partner reagents, whereas homopolymers are depicted as formal “heteropolymers” with identical partners. The key rule proposed here is to choose “formal” monomers such as to minimize the number of marked atoms, involved in bonds being formed or changing bond order. Accordingly, carbonyl compounds are rendered as the less-stable vinyl alcohol tautomer, following the same formalism as in olefin polymerization, to minimize the total number of formal polymerization mechanisms and herewith provide the most general framework encompassing a maximum of polymerization processes. ISIDA (in silico design and data analysis) fragment counts with special status given to the “marked atoms” participating in the polymerization process were combined using “mixture” strategies to generate the final polymer descriptors. Three predictive models based on SVR (support vector regression) are discussed here. After reproducing results of Katritzky et al. with a local model applicable only to linear homo/heteropolymers, an epoxy resin-specific model applicable to both linear and network forms was built. Eventually, the general model applicable to all these families was constructed. In 12 × repeated 3-fold cross-validation challenges, it displayed the highest accuracy of Q2 = 0.920, RMSE = 34.3 K over the training set of 270 polymers, and R2 = 0.779, RMSE 35.9 K for an external test set of 119 polymers. GTM (Generative Topographic Mapping) analysis produced a 2D map of “polymer chemical space”, highlighting the various classes of polymers included in the study and their relationship with respect to Tg values. The epoxy-specific and general models are publicly available on our web server: http://infochim.u-strasbg.fr/webserv/VSEngine.html.

Combination of envelope spectra and generative topographic mapping to diagnose bearing fault and evaluate degradation of bearing

Functioning of rotary machines is based on the rolling element of bearings, as the rolling element of bearings is the crucial component of rotary machines. The entire operation of rotating machines may terminate after the failure of rolling element of bearings. Hence, the failure of rotating machine is mainly dependent on rolling element of bearings, as it may break down an entire process. Thus, to avoid the sudden breakdown of rotary machines, it is essential to develop the diagnostics and prognostics methodology of rolling element of bearings. In the rolling element of bearings, the assessment of bearing degradation and fault diagnostics is essential to establish the methodology. Therefore, this article proposes a methodology to diagnose the fault of bearing and to assess bearing degradation. Envelope spectra and generative topographic mapping are used together in the methodology proposed. Selection of frequency band based on encompasses one or more resonance of spectrum of signal. Envelope spectrum is determined to pick up the peak of the characteristic fault frequencies of the bearing. Extraction of features from time and frequency domains is used in the generative topographic mapping. Health degradation index, an index termed as evolution of bearing degradation, is obtained from the classifier of generative topographic mapping. Experiments were conducted on the bearings to verify the proposed methodology. On interpreting the obtained results, it was found that the proficiency of proposed methodology is most appropriate for detecting the fault in bearing and tracking the degradation evolution in bearings.

Probabilistic ancestry maps: a method to assess and visualize population substructures in genetics

BackgroundPrincipal component analysis (PCA) is a standard method to correct for population stratification in ancestry-specific genome-wide association studies (GWASs) and is used to cluster individuals by ancestry. Using the 1000 genomes project data, we examine how non-linear dimensionality reduction methods such as t-distributed stochastic neighbor embedding (t-SNE) or generative topographic mapping (GTM) can be used to provide improved ancestry maps by accounting for a higher percentage of explained variance in ancestry, and how they can help to estimate the number of principal components necessary to account for population stratification. GTM generates posterior probabilities of class membership which can be used to assess the probability of an individual to belong to a given population - as opposed to t-SNE, GTM can be used for both clustering and classification.ResultsPCA only partially identifies population clusters and does not separate most populations within a given continent, such as Japanese and Han Chinese in East Asia, or Mende and Yoruba in Africa. t-SNE and GTM, taking into account more data variance, can identify more fine-grained population clusters. GTM can be used to build probabilistic classification models, and is as efficient as support vector machine (SVM) for classifying 1000 Genomes Project populations.ConclusionThe main interest of probabilistic GTM maps is to attain two objectives with only one map: provide a better visualization that separates populations efficiently, and infer genetic ancestry for individuals or populations. This paper is a first application of GTM for ancestry classification models. Our code (https://github.com/hagax8/ancestry_viz) and interactive visualizations (https://lovingscience.com/ancestries) are available online.

De Novo Molecular Design by Combining Deep Autoencoder Recurrent Neural Networks with Generative Topographic Mapping.

Here we show that Generative Topographic Mapping (GTM) can be used to explore the latent space of the SMILES-based autoencoders and generate focused molecular libraries of interest. We have built a sequence-to-sequence neural network with Bidirectional Long Short-Term Memory layers and trained it on the SMILES strings from ChEMBL23. Very high reconstruction rates of the test set molecules were achieved (>98%), which are comparable to the ones reported in related publications. Using GTM, we have visualized the autoencoder latent space on the two-dimensional topographic map. Targeted map zones can be used for generating novel molecular structures by sampling associated latent space points and decoding them to SMILES. The sampling method based on a genetic algorithm was introduced to optimize compound properties "on the fly". The generated focused molecular libraries were shown to contain original and a priori feasible compounds which, pending actual synthesis and testing, showed encouraging behavior in independent structure-based affinity estimation procedures (pharmacophore matching, docking).

Getting to Know the Neighbours with GTM: The Case of Antiviral Compounds.

Recent outbreaks of dangerous viral infections, such as Ebola virus disease, Zika fever, etc., are forcing the search for new antiviral compounds. Preferably, such compounds should possess broad-spectrum antiviral activity, as the development of drugs for the treatment of dozens of viral infections lacking specific treatment would require significant resources. Antiviral activity data present in public resources are very sparse and further investigation of structure-activity relationships is necessary. One of the strategies could be the investigation of chemical space around known active compounds and assessment of activity against closely related viruses in order to fill in the antiviral activity matrix. Here we present an investigation of antiviral activity using universal maps built with generative topographic mapping (GTM) algorithm. The GTM-based maps were used to find commercially available compounds in close proximity to already known compounds with anti-flaviviral and anti-enteroviral activities. Selected compounds were then assessed in cell-based assays against tick-borne encephalitis virus (TBEV) and a panel of enteroviruses. This approach allowed us to identify 23 new compounds showing anti-TBEV activity with EC50 values in micromolar and submicromolar range.

Multi-task generative topographic mapping in virtual screening.

The previously reported procedure to generate "universal" Generative Topographic Maps (GTMs) of the drug-like chemical space is in practice a multi-task learning process, in which both operational GTM parameters (example: map grid size) and hyperparameters (key example: the molecular descriptor space to be used) are being chosen by an evolutionary process in order to fit/select "universal" GTM manifolds. After selection (a one-time task aimed at optimizing the compromise in terms of neighborhood behavior compliance, over a large pool of various biological targets), for any further use the manifolds are ready to provide "fit-free" predictive models. Using any structure-activity set-irrespectively whether the associated target served at map fitting stage or not-the generation or "coloring" a property landscape enables predicting the property for any external molecule, with zero additional fitable parameters involved. While previous works have signaled the excellent behavior of such models in aggressive three-fold cross-validation assessments of their predictive power, the present work wished to explore their behavior in Virtual Screening (VS), here simulated on hand of external DUD ligand and decoy series that are fully disjoint from the ChEMBL-extracted landscape coloring sets. Beyond the rather robust results of the universal GTM manifolds in this challenge, it could be shown that the descriptor spaces selected by the evolutionary multi-task learner were intrinsically able to serve as an excellent support for many other VS procedures, starting from parameter-free similarity searching, to local (target-specific) GTM models, to parameter-rich, nonlinear Random Forest and Neural Network approaches.

SA23 - BIOLOGICAL PATHWAYS AND DRUG GENE-SETS: ANALYSIS AND VISUALIZATION

Background Here, we use summary statistics from Genome-Wide Association Studies (GWAS) to find significant biological pathways, disease pathways, drug targets and drug gene-sets in schizophrenia, BMI, Body Fat % (BF%), and Fat Free Mass (FFM). We introduce a workflow encompassing data collection and curation, analysis of drug gene-sets and biological pathways, drug class enrichment analysis, and visualization of pathway landscapes. Methods We used the schizophrenia GWAS from the PGC Schizophrenia working group phase 2 (35,476 cases), and new GWASs of BMI, BF% and FFM based on UK Biobank data (N=83,477). We curated drug/gene interactions from various databases: DGIdb, DSigDB, Ki DB, PHAROS, and ChEMBL. Biological pathways were collected from MSigDB (GO and canonical pathways) and disease/phenotype pathways from the Open Targets platform. The software MAGMA was used to produce gene-wise and pathway-wise associations, and MetaXcan for transcriptome prediction based on GTEx data (Genotype-Tissue Expression project). The enrichment of drug classes was estimated by grouping drugs by ATC class (Anatomical Therapeutic Chemical) and estimating the enrichment using Wilcoxon-Mann-Whitney test and the AUC (area under the enrichment curve). Pathway landscapes were generated using the kernel Generative Topographic Mapping (GTM), a probabilistic dimensionality reduction algorithm. On these maps, gene-sets are represented by points and the background color is derived from GWAS analysis p-values. Results Results show new Bonferroni-significant pathways in schizophrenia. Antipsychotics and antiepileptics are enriched in the latest and largest schizophrenia GWAS from the PGC Schizophrenia working group. Different disease/phenotype pathways are found in BMI and BF% - e.g., “binge eating” and “age at menarche” for BMI. 47 disease/phenotype pathways and 24 biological pathways are significantly associated with FFM: anthropometric measures, height, myopathies, midgut development, bone development, etc. Pathway landscapes reveal that many significant pathways overlap in “pathway clusters”, such as the skeletal system cluster or the myopathy cluster on FFM maps. Discussion A comprehensive approach is necessary to investigate drug repurposing opportunities and biological pathways using GWAS results. New visualization approaches may help to highlight the overlap between pathways and the genes driving the association. Two key issues are data availability (drug/target affinities, gene annotations) and sample sizes for GWAS studies. Growing public databases and increasing sample sizes will help us to improve our understanding of the genetic etiology of complex diseases.