Receptor-based Methods Research Articles

Preface to the special issue in memory of Phil Magee

Shortly after Phil Magee died of cancer in June 2005, Marvin Charton approached me at the end of an AGRO Division QSAR Symposium and asked me if I would be willing to help him organize a memorial Symposium in Phil’s honor. I was happy to do so, and was pleased to be joined in the effort by John Block the next June. Lowell Hall and Lemont Kier offered to pitch in later that year. Sponsoring organizations eventually included the AGRO, COMP and CINF Divisions of the American Chemical Society and the Cheminformatics and QSAR Society. Corporate sponsorship was provided by this Journal, Tripos International, Chemical Computing Group, eduSoft, ChemSilico, BioSAR Research, Hall Associates Consulting and Magee Real Estate. Their generous contributions were used to offset the travel expenses of several Symposium speakers who would not have been able participate otherwise. The response to our calls for Symposium papers was nearly overwhelming. Over 40 abstracts for oral presentations were submitted, with several of Phil’s friends and colleagues writing to express their regret at not being able to attend the meeting. The memorial symposium, formally titled ‘‘QSAR Reborn: A Symposium in Memory of Dr. Phillip Magee’’, consisted of six well-attended sessions spread across all five days (19–23 August 2007) of the 234th National ACS Meeting in Boston MA. The idea of putting together the Memorial issue of the Journal of Computer-Aided Molecular Design that you now hold in your hands—physically or virtually—grew in parallel to the organization of the Symposium. Symposium participants were invited to contribute, of course, but the topics were not restricted to what had been presented at the meeting. The call for papers solicited manuscripts for the Journal from the entire community, including those who were unable to attend or to speak at the Boston meeting. Just as for the Symposium, the response for Journal contributions was more than could reasonably have been hoped for. I trust that you will find the 15 papers collected here a fitting tribute to Phil’s memory. The authors hail from across the continent and around the world, representing ten states and provinces and nine different countries. They are drawn from industrial as well as academic research organizations. The collection begins with Marvin Charton’s brief biography of the man and his place in QSAR and its history. About half of the technical papers deal with 2D descriptors and about half involve 3D (or higher) levels of description, with some comparing—or even integrating— the two levels. Topics range from the practical (e.g., predicting solubility or developing a better insecticide) to the esoteric (e.g., scaling descriptors by molecular size to decorrelate them or choosing between ligandand receptorbased alignment methods). Most deal with QSAR at the level of the intact organism, just as Phil himself often did. The organisms considered range in size from viruses to guppies to humans. Although in vitro assays of one kind or another underlie some of the analyses, most concern physical, metabolic or toxicological properties. They include two studies on dermal sensitivity, one of which makes use of a data set compiled by Phil himself. In summary: the papers collected here embrace nearly as broad a chunk of the QSAR world as Phil Magee did himself. I think he would approve, and hope that you will as well.

Assessing Sources of PM2.5 in Cities Influenced by Regional Transport

The human health effects of fine particulate matter (PM2.5) have provided impetus for the establishment of new air quality standards or guidelines in many countries. This has led to the need for information on the main sources responsible for PM2.5. In urban locations being impacted by regional-scale transport, source-receptor relationships for PM2.5 are complex and require the application of multiple receptor-based analysis methods to gain a better understanding. This approach is being followed to study the sources of PM2.5 impacting southern Ontario, Canada, and its major city of Toronto. Existing monitoring data in the region around Toronto and within Toronto itself are utilized to estimate that 30–45% of the PM2.5 is from local sources, which implies that 55–70% is transported into the area. In addition, there are locations in the city that can be shown to experience a greater impact from local sources such as motor vehicle traffic. Detailed PM2.5 chemical characterization data were collected in Toronto in order to apply two different multivariate receptor models to determine the main sources of the PM2.5. Both approaches produced similar results, indicating that motor-vehicle-related emissions, most likely of local origin, are directly responsible for about 20% of the PM2.5. Gasoline engine vehicles were found to be a greater overall contributor (13%) compared to diesel vehicles (8%). Secondary PM2.5 from coal-fired power plants continues to be a significant contributor (20–25%) and also played a role in enhancing production of secondary organic carbon mass (15%) on fine particles. Secondary fine particle nitrate was the single most important source (35%), with a large fraction of this likely related to motor vehicle emissions. Independent use of different receptor models helps provide more confidence in the source apportionment, as does comparison of results among complementary receptor-based data analysis approaches.

ProPose: Steered Virtual Screening by Simultaneous Protein—Ligand Docking and Ligand—Ligand Alignment.

The ‘model-free' screening engine ProPose implements a general method for performing simultaneous protein−ligand docking, ligand−ligand alignment, pharmacophore queriesand combinations thereofin order to incorporate a priori information into screening protocols. In this manuscript we describe a case study on herpes simplex virus thymidine kinase, an important antiviral drug target, where we evaluate different approaches for handling a specific type of a priori information, i.e., multiple target structures. We demonstrate that a simultaneous alignment on two target structuresin conjunction with logic operations on interactions and docking constraints derived from protein structureis an effective means of (i) improving the enrichment of chemical substructures that are compatible with the a priori known ligands, (ii) ensuring the steric fit into the target protein, and (iii) handling target flexibility. The combination of ligand- and receptor-based methods steers the virtual screening by ranking molecules ac...

ProPose: Steered Virtual Screening by Simultaneous Protein−Ligand Docking and Ligand−Ligand Alignment

The 'model-free' screening engine ProPose implements a general method for performing simultaneous protein-ligand docking, ligand-ligand alignment, pharmacophore queries-and combinations thereof-in order to incorporate a priori information into screening protocols. In this manuscript we describe a case study on herpes simplex virus thymidine kinase, an important antiviral drug target, where we evaluate different approaches for handling a specific type of a priori information, i.e., multiple target structures. We demonstrate that a simultaneous alignment on two target structures--in conjunction with logic operations on interactions and docking constraints derived from protein structure--is an effective means of (i) improving the enrichment of chemical substructures that are compatible with the a priori known ligands, (ii) ensuring the steric fit into the target protein, and (iii) handling target flexibility. The combination of ligand- and receptor-based methods steers the virtual screening by ranking molecules according to the similarity of their interaction pattern with known ligands, thereby--to some extent--outweighing the deficiencies of simple scoring functions often used in initial virtual screening.

Protein flexibility is an important component of structure-based drug discovery.

Receptor-based drug discovery can increase the novelty of a hit list over ligand-based models that are dependent on known inhibitors. It is important to explore new conformational and chemical space, but it is difficult to predict the plasticity of the binding site. Receptor-based methods are usually based on crystal structures of ligand-protein complexes, and hit lists can be restricted to the size and shape of the receptor model. Many improvements that accommodate protein flexibility in computer-aided drug design are being developed. These methods are reviewed with the focus being techniques that move beyond the rotation of side chains. The use of multiple protein structures is emerging as the best choice for including more realistic changes in protein conformation, but the optimal way to using these structures is still unclear.

Safety of dimethoate insecticide.

<h3>Abstract</h3> 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 machine-learning 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.