Predicting target proteins for drug candidate compounds based on drug-induced gene expression data in a chemical structure-independent manner

Hizukuri, Yoshiyuki; Sawada, Ryusuke; Yamanishi, Yoshihiro

doi:10.1186/s12920-015-0158-1

Cited by 38 publications

(29 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Various chemogenomics methods have been proposed in the last decade [13,14,15,16,17,18,19,20,21,22,23,24,25,26,11]. They differ by (i) the descriptors used to encode proteins and ligands or the similarities between these objects, (ii) the ML algorithm that is used to learn the model and make the predictions.…”

Section: Pioneering Work In Machine-learning For Chemogenomicsmentioning

confidence: 99%

Evaluation of network architecture and data augmentation methods for deep learning in chemogenomics

Playe

Stoven

2019

Preprint

View full text Add to dashboard Cite

Among virtual screening methods that have been developed to facilitate the drug discovery process, chemogenomics presents the particularity to tackle the question of predicting ligands for proteins, at at scales both in the protein and chemical spaces. Therefore, in addition to to predict drug candidates for a given therapeutic protein target, like more classical ligand-based or receptor-based methods do, chemogenomics can also predict off-targets at the proteome level, and therefore, identify potential side-effects or drug repositioning opportunities. In this study, we study and compare machinelearning and deep learning approaches for chemogenomics, that are applicable to screen large sets of compounds against large sets of druggable proteins. State-of-the-art drug chemogenomics methods rely on expert-based chemical and protein descriptors or similarity measures. The recent development of deep learning approaches enabled to design algorithms that learn numerical abstract representations of molecular graphs and protein sequences in an end-to-end fashion, i.e., so that the learnt features optimise the objective function of the drug-target interaction prediction task. In this paper, we address drug-target interaction prediction at the druggable proteome-level, with what we define as the chemogenomic neuron network. This network consists of a feed-forward neuron network taking as input the combination of molecular and protein representations learnt by molecular graph and protein sequence encoders. We first propose a standard formulation of this chemogenomic neuron network. Then, we compare the performances of the standard chemogenomic network to reference deep learning or shallow (machine-learning without deep learning) methods. In particular, we show that such a representation learning approach is competitive with state-of-the-art chemogenomics with shallow methods, but not ultimately superior. We evaluate the most promising neuron network architectures and data augmentation techniques, such as multi-view and transfer learning, to improve the prediction performance of the chemogenomic network. Our results shed new insights on the design of chemogenomics approaches based on representation learning algorithms. Most importantly, we conclude from our observations that a promising research direction is to integrate heterogeneous sources of data such as various bioactivity datasets, or independently, multiple molecule and protein attribute views, instead of focusing on sophisticated, yet intuitively relevant, encoder's neuron network architecture.

show abstract

Section: Pioneering Work In Machine-learning For Chemogenomicsmentioning

confidence: 99%

Evaluation of network architecture and data augmentation methods for deep learning in chemogenomics

Playe

Stoven

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Various chemogenomics methods have been proposed in the last decade. [14][15][16][17][18][19][20][21][22] They all rely on the assumption that "similar" proteins are expected to bind "similar" ligands. These methods can be further divided into two categories, known as Single-Task (ST) and Multi-Task (MT) in the ML literature.…”

Section: Drug Specificitymentioning

confidence: 99%

“…All work directly inspired from their pioneering work use this algorithm as a classifier. [16][17][18][19][20]38 In the Supporting Information, we briefly recall the basic principles of these methods.…”

Section: Kernel-based Classifiersmentioning

confidence: 99%

Efficient multi-task chemogenomics for drug specificity prediction

Playe

Azencott

Stoven

2017

Preprint

View full text Add to dashboard Cite

Adverse drug reactions (also called side effects) often interrupt the long and costly process of drug development. Side effects occur when drugs bind to proteins other than their intended target. As experimentally testing drug specificity against the entire proteome is out of reach, we investigate in this paper the application of chemogenomics approaches.We formulate the study of drug specificity as a proteome-wide drug-protein interaction prediction problem, and build appropriate data sets on which we evaluate multi-task support vector machines.Our observations lead us to propose NNMT, an efficient Multi-Task SVM for chemogenomics that is trained on a limited number of data points. Finally, we demonstrate the suitability of NNMT to study the specificity of drug-like molecules in real-life situations by suggesting secondary targets for 36 recently withdrawn drugs. In 9 cases, we identified secondary targets responsible for the withdrawal of the drug.

show abstract

“…Therefore, the data predicted in in silico approaches would be recorded into HEDD, such as molecular docking, virtual screening, pharmacophore modeling, molecular dynamics and similarity searching. As a resource to study the potential roles of epigenetic drugs in remodeling epigenetic modification, HEDD could be extended with utilities for the identification and confirmation of targets (genes and pathways) related to epigenetic drugs from large-scale high-throughput data (such as gene expression) (42, 43). Since mice are very important for modeling diseases and testing drugs, we would extend the research scope and integrate high-throughput data for mice treated with epigenetic drugs into HEDD.…”

Section: Database Use and Accessmentioning

confidence: 99%

HEDD: the human epigenetic drug database

Qi¹,

Wang²,

Wang³

et al. 2016

Database

View full text Add to dashboard Cite

Epigenetic drugs are chemical compounds that target disordered post-translational modification of histone proteins and DNA through enzymes, and the recognition of these changes by adaptor proteins. Epigenetic drug-related experimental data such as gene expression probed by high-throughput sequencing, co-crystal structure probed by X-RAY diffraction and binding constants probed by bio-assay have become widely available. The mining and integration of multiple kinds of data can be beneficial to drug discovery and drug repurposing. HEMD and other epigenetic databases store comprehensively epigenetic data where users can acquire segmental information of epigenetic drugs. However, some data types such as high-throughput datasets are not provide by these databases and they do not support flexible queries for epigenetic drug-related experimental data. Therefore, in reference to HEMD and other epigenetic databases, we developed a relatively comprehensive database for human epigenetic drugs. The human epigenetic drug database (HEDD) focuses on the storage and integration of epigenetic drug datasets obtained from laboratory experiments and manually curated information. The latest release of HEDD incorporates five kinds of datasets: (i) drug, (ii) target, (iii) disease, (vi) high-throughput and (v) complex. In order to facilitate data extraction, flexible search options were built in HEDD, which allowed an unlimited condition query for specific kinds of datasets using drug names, diseases and experiment types.Database URL: http://hedds.org/

show abstract

Predicting target proteins for drug candidate compounds based on drug-induced gene expression data in a chemical structure-independent manner

Cited by 38 publications

References 39 publications

Evaluation of network architecture and data augmentation methods for deep learning in chemogenomics

Evaluation of network architecture and data augmentation methods for deep learning in chemogenomics

Efficient multi-task chemogenomics for drug specificity prediction

HEDD: the human epigenetic drug database

Contact Info

Product

Resources

About