“Virtual Fragment Linking”: An Approach To Identify Potent Binders from Low Affinity Fragment Hits

DRlinker: Deep Reinforcement Learning for Optimization in Fragment Linking Design

Fragment-based drug discovery is a widely used strategy for drug design in both academic and pharmaceutical industries. Although fragments can be linked to generate candidate compounds by the latest deep generative models, generating linkers with specified attributes remains underdeveloped. In this study, we presented a novel framework, DRlinker, to control fragment linking toward compounds with given attributes through reinforcement learning. The method has been shown to be effective for many tasks from controlling the linker length and log P, optimizing predicted bioactivity of compounds, to various multiobjective tasks. Specifically, our model successfully generated 91.0% and 93.9% of compounds complying with the desired linker length and log P and improved the 7.5 pChEMBL value in bioactivity optimization. Finally, a quasi-scaffold-hopping study revealed that DRlinker could generate nearly 30% molecules with high 3D similarity but low 2D similarity to the lead inhibitor, demonstrating the benefits and applicability of DRlinker in actual fragment-based drug design.

SyntaLinker: automatic fragment linking with deep conditional transformer neural networks

Linking fragments to generate a focused compound library for a specific drug target is one of the challenges in fragment-based drug design (FBDD). Hereby, we propose a new program named SyntaLinker, which is based on a syntactic pattern recognition approach using deep conditional transformer neural networks. This state-of-the-art transformer can link molecular fragments automatically by learning from the knowledge of structures in medicinal chemistry databases (e.g. ChEMBL database). Conventionally, linking molecular fragments was viewed as connecting substructures that were predefined by empirical rules. In SyntaLinker, however, the rules of linking fragments can be learned implicitly from known chemical structures by recognizing syntactic patterns embedded in SMILES notations. With deep conditional transformer neural networks, SyntaLinker can generate molecular structures based on a given pair of fragments and additional restrictions. Case studies have demonstrated the advantages and usefulness of SyntaLinker in FBDD.

Pharmacophore modeling, 3D-QSAR and molecular docking studies of quinazolines and aminopyridines as selective inhibitors of inducible nitric oxide synthase

Pharmacophore modeling and 3D-Quantitative Structure Activity Relationship (3D-QSAR) studies have been performed on a dataset of thirty-two quinazoline and aminopyridine derivatives to get an insight into the important structural features required for binding to inducible nitric oxide synthase (iNOS). A four-point CPH (Common Pharmacophore Hypothesis), AHPR.29, with a hydrogen bond acceptor, hydrophobic group, positively charged ionizable group and an aromatic ring, has been obtained as the best pharmacophore model. Satisfactory statistical parameters of correlation ([Formula: see text]) and cross-validated ([Formula: see text]) correlation coefficients, 0.9288 and 0.6353, respectively, show high robustness and good predictive ability of our selected model. The contour maps have been developed from this model and the analysis has provided an interpretable explanation of the effect that various features and substituents have on the potency and selectivity of inhibitors towards iNOS. Docking studies have also been performed in order to analyze the interactions between the enzyme and the inhibitors. Our proposed model can thus be further used for screening a large database of compounds and design new iNOS inhibitors.

Impact of a Central Scaffold on the Binding Affinity of Fragment Pairs Isolated from DNA-Encoded Self-Assembling Chemical Libraries

The screening of encoded self-assembling chemical libraries allows the identification of fragment pairs that bind to adjacent pockets on target proteins of interest. For practical applications, it is necessary to link these ligand pairs into discrete organic molecules, devoid of any nucleic acid component. Here we describe the discovery of a synergistic binding pair for acid alpha-1 glycoprotein and a chemical strategy for the identification of optimal linkers, connecting the two fragments. The procedure yielded a set of small organic ligands, the best of which exhibited a dissociation constant of 9.9 nm, as measured in solution by fluorescence polarization.

Bayesian models trained with HTS data for predicting β-haematin inhibition and in vitro antimalarial activity

A large quantity of high throughput screening (HTS) data for antimalarial activity has become available in recent years. This includes both phenotypic and target-based activity. Realising the maximum value of these data remains a challenge. In this respect, methods that allow such data to be used for virtual screening maximise efficiency and reduce costs. In this study both in vitro antimalarial activity and inhibitory data for β-haematin formation, largely obtained from publically available sources, has been used to develop Bayesian models for inhibitors of β-haematin formation and in vitro antimalarial activity. These models were used to screen two in silico compound libraries. In the first, the 1510 U.S. Food and Drug Administration approved drugs available on PubChem were ranked from highest to lowest Bayesian score based on a training set of β-haematin inhibiting compounds active against Plasmodium falciparum that did not include any of the clinical antimalarials or close analogues. The six known clinical antimalarials that inhibit β-haematin formation were ranked in the top 2.1% of compounds. Furthermore, the in vitro antimalarial hit-rate for this prioritised set of compounds was found to be 81% in the case of the subset where activity data are available in PubChem. In the second, a library of about 5000 commercially available compounds (Aldrich(CPR)) was virtually screened for ability to inhibit β-haematin formation and then for in vitro antimalarial activity. A selection of 34 compounds was purchased and tested, of which 24 were predicted to be β-haematin inhibitors. The hit rate for inhibition of β-haematin formation was found to be 25% and a third of these were active against P. falciparum, corresponding to enrichments estimated at about 25- and 140-fold relative to random screening, respectively.

Combinatorial Consensus Scoring for Ligand-Based Virtual Fragment Screening: A Comparative Case Study for Serotonin 5-HT(3)A, Histamine H(1), and Histamine H(4) Receptors

In the current study we have evaluated the applicability of ligand-based virtual screening (LBVS) methods for the identification of small fragment-like biologically active molecules using different similarity descriptors and different consensus scoring approaches. For this purpose, we have evaluated the performance of 14 chemical similarity descriptors in retrospective virtual screening studies to discriminate fragment-like ligands of three membrane-bound receptors from fragments that are experimentally determined to have no affinity for these proteins (true inactives). We used a complete fragment affinity data set of experimentally determined ligands and inactives for two G protein-coupled receptors (GPCRs), the histamine H1 receptor (H1R) and the histamine H4 receptor (H4R), and one ligand-gated ion channel (LGIC), the serotonin receptor (5-HT3AR), to validate our retrospective virtual screening studies. We have exhaustively tested consensus scoring strategies that combine the results of multiple actives (group fusion) or combine different similarity descriptors (similarity fusion), and for the first time systematically evaluated different combinations of group fusion and similarity fusion approaches. Our studies show that for these three case study protein targets both consensus scoring approaches can increase virtual screening enrichments compared to single chemical similarity search methods. Our cheminformatics analyses recommend to use a combination of both group fusion and similarity fusion for prospective ligand-based virtual fragment screening.

Fragment library design: using cheminformatics and expert chemists to fill gaps in existing fragment libraries

Fragment based screening (FBS) has emerged as a mainstream lead discovery strategy in academia, biotechnology start-ups, and large pharma. As a prerequisite of FBS, a structurally diverse library of fragments is desirable in order to identify chemical matter that will interact with the range of diverse target classes that are prosecuted in contemporary screening campaigns. In addition, it is also desirable to offer synthetically amenable starting points to increase the probability of a successful fragment evolution through medicinal chemistry. Herein we describe a method to identify biologically relevant chemical substructures that are missing from an existing fragment library (chemical gaps), and organize these chemical gaps hierarchically so that medicinal chemists can efficiently navigate the prioritized chemical space and subsequently select purchasable fragments for inclusion in an enhanced fragment library.

Biofragments: an approach towards predicting protein function using biologically related fragments and its application to Mycobacterium tuberculosis CYP126

We present a novel fragment-based approach that tackles some of the challenges for chemical biology of predicting protein function. The general approach, which we have termed biofragments, comprises two key stages. First, a biologically relevant fragment library (biofragment library) can be designed and constructed from known sets of substrate-like ligands for a protein class of interest. Second, the library can be screened for binding to a novel putative ligand-binding protein from the same or similar class, and the characterization of hits provides insight into the basis of ligand recognition, selectivity, and function at the substrate level. As a proof-of-concept, we applied the biofragments approach to the functionally uncharacterized Mycobacterium tuberculosis (Mtb) cytochrome P450 isoform, CYP126. This led to the development of a tailored CYP biofragment library with notable 3D characteristics and a significantly higher screening hit rate (14%) than standard drug-like fragment libraries screened previously against Mtb CYP121 and 125 (4% and 1%, respectively). Biofragment hits were identified that make both substrate-like type-I and inhibitor-like type-II interactions with CYP126. A chemical-fingerprint-based substrate model was built from the hits and used to search a virtual TB metabolome, which led to the discovery that CYP126 has a strong preference for the recognition of aromatics and substrate-like type-I binding of chlorophenol moieties within the active site near the heme. Future catalytic analyses will be focused on assessing CYP126 for potential substrate oxidative dehalogenation.

Efficient search of chemical space: navigating from fragments to structurally diverse chemotypes

We introduce a novel strategy to sample bioactive chemical space, which follows-up on hits from fragment campaigns without the need for a crystal structure. Our results strongly suggest that screening a few hundred or thousand fragments can substantially improve the selection of small-molecule screening subsets. By combining fragment-based screening with virtual fragment linking and HTS fingerprints, we have developed an effective strategy not only to expand from low-affinity hits to potent compounds but also to hop in chemical space to substantially novel chemotypes. In benchmark calculations, our approach accessed subsets of compounds that were substantially enriched in chemically diverse hit compounds for various activity classes. Overall, half of the hits in the screening collection were found by screening only 10% of the library. Furthermore, a prospective application led to the discovery of two structurally novel histone deacetylase 4 inhibitors.

In silico target predictions: defining a benchmarking data set and comparison of performance of the multiclass Naïve Bayes and Parzen-Rosenblatt window

In this study, two probabilistic machine-learning algorithms were compared for in silico target prediction of bioactive molecules, namely the well-established Laplacian-modified Naïve Bayes classifier (NB) and the more recently introduced (to Cheminformatics) Parzen-Rosenblatt Window. Both classifiers were trained in conjunction with circular fingerprints on a large data set of bioactive compounds extracted from ChEMBL, covering 894 human protein targets with more than 155,000 ligand-protein pairs. This data set is also provided as a benchmark data set for future target prediction methods due to its size as well as the number of bioactivity classes it contains. In addition to evaluating the methods, different performance measures were explored. This is not as straightforward as in binary classification settings, due to the number of classes, the possibility of multiple class memberships, and the need to translate model scores into "yes/no" predictions for assessing model performance. Both algorithms achieved a recall of correct targets that exceeds 80% in the top 1% of predictions. Performance depends significantly on the underlying diversity and size of a given class of bioactive compounds, with small classes and low structural similarity affecting both algorithms to different degrees. When tested on an external test set extracted from WOMBAT covering more than 500 targets by excluding all compounds with Tanimoto similarity above 0.8 to compounds from the ChEMBL data set, the current methodologies achieved a recall of 63.3% and 66.6% among the top 1% for Naïve Bayes and Parzen-Rosenblatt Window, respectively. While those numbers seem to indicate lower performance, they are also more realistic for settings where protein targets need to be established for novel chemical substances.

Design and synthesis of novel lactate dehydrogenase A inhibitors by fragment-based lead generation

Lactate dehydrogenase A (LDHA) catalyzes the conversion of pyruvate to lactate, utilizing NADH as a cofactor. It has been identified as a potential therapeutic target in the area of cancer metabolism. In this manuscript we report our progress using fragment-based lead generation (FBLG), assisted by X-ray crystallography to develop small molecule LDHA inhibitors. Fragment hits were identified through NMR and SPR screening and optimized into lead compounds with nanomolar binding affinities via fragment linking. Also reported is their modification into cellular active compounds suitable for target validation work.

Emerging role of surface plasmon resonance in fragment-based drug discovery

Surface plasmon resonance (SPR) offers a method of biophysical fragment screening that is fast, efficient, cost effective and accurate. SPR is increasingly being adopted as a secondary assay to validate fragment hits. Recently, technical advances have resulted in the emergence of SPR as a primary screening methodology for fragment-based drug discovery. Moreover, SPR biosensor assays can be developed for a wide range of proteins, including membrane proteins, such as G-protein-coupled receptors. In this review, we discuss the advantages and limitations of SPR fragment screening including experimental consideration of reducing false positive and false negative rates to a minimum. We discuss how ligand efficiency can be used both as a method to eliminate false positives and to understand which fragments in a library may be a source of false negatives.

Fragment-based approaches and computer-aided drug discovery

Fragment-based design has significantly modified drug discovery strategies and paradigms in the last decade. Besides technological advances and novel therapeutic avenues, one of the most significant changes brought by this new discipline has occurred in the minds of drug designers. Fragment-based approaches have markedly impacted rational computer-aided design both in method development and in applications. The present review illustrates the importance of molecular fragments in many aspects of rational ligand design, and discusses how thinking in "fragment space" has boosted computational biology and chemistry.

Bayesian methods in virtual screening and chemical biology

The Naïve Bayesian Classifier, as well as related classification and regression approaches based on Bayes' theorem, has experienced increased attention in the cheminformatics world in recent years. In this contribution, we first review the mathematical framework on which Bayes' methods are built, and then continue to discuss implications of this framework as well as practical experience under which conditions Bayes' methods give the best performance in virtual screening settings. Finally, we present an overview of applications of Bayes' methods to both virtual screening and the chemical biology arena, where applications range from bridging phenotypic and mechanistic space of drug action to the prediction of ligand-target interactions.

The multiple roles of computational chemistry in fragment-based drug design

Fragment-based drug discovery (FBDD) represents a change in strategy from the screening of molecules with higher molecular weights and physical properties more akin to fully drug-like compounds, to the screening of smaller, less complex molecules. This is because it has been recognised that fragment hit molecules can be efficiently grown and optimised into leads, particularly after the binding mode to the target protein has been first determined by 3D structural elucidation, e.g. by NMR or X-ray crystallography. Several studies have shown that medicinal chemistry optimisation of an already drug-like hit or lead compound can result in a final compound with too high molecular weight and lipophilicity. The evolution of a lower molecular weight fragment hit therefore represents an attractive alternative approach to optimisation as it allows better control of compound properties. Computational chemistry can play an important role both prior to a fragment screen, in producing a target focussed fragment library, and post-screening in the evolution of a drug-like molecule from a fragment hit, both with and without the available fragment-target co-complex structure. We will review many of the current developments in the area and illustrate with some recent examples from successful FBDD discovery projects that we have conducted.

Formal concept analysis for the identification of molecular fragment combinations specific for active and highly potent compounds

We introduce fragment formal concept analysis (FragFCA) to study complex relationships between fragments in active compounds taking potency information into account. Fragment combinations that are unique to active or highly potent compounds or that are shared by molecules having different or overlapping activity profiles are systematically identified using chemically intuitive queries of varying complexity. The methodology is applied to analyze fragment distributions in antagonists of seven G protein coupled receptor targets and identify signature fragments. Pairs or triplets of molecular fragments are found to be most specific for different activity profiles and compound potency levels. In addition, we demonstrate the ability of FragFCA to identify selective hits in high-throughput screening data sets.