Dávid Bajusz

Why is Tanimoto index an appropriate choice for fingerprint-based similarity calculations?

AbstractBackgroundCheminformaticians are equipped with a very rich toolbox when carrying out molecular similarity calculations. A large number of molecular representations exist, and there are several methods (similarity and distance metrics) to quantify the similarity of molecular representations. In this work, eight well-known similarity/distance metrics are compared on a large dataset of molecular fingerprints with sum of ranking differences (SRD) and ANOVA analysis. The effects of molecular size, selection methods and data pretreatment methods on the outcome of the comparison are also assessed.ResultsA supplier database (https://mcule.com/) was used as the source of compounds for the similarity calculations in this study. A large number of datasets, each consisting of one hundred compounds, were compiled, molecular fingerprints were generated and similarity values between a randomly chosen reference compound and the rest were calculated for each dataset. Similarity metrics were compared based on their ranking of the compounds within one experiment (one dataset) using sum of ranking differences (SRD), while the results of the entire set of experiments were summarized on box and whisker plots. Finally, the effects of various factors (data pretreatment, molecule size, selection method) were evaluated with analysis of variance (ANOVA).ConclusionsThis study complements previous efforts to examine and rank various metrics for molecular similarity calculations. Here, however, an entirely general approach was taken to neglect any a priori knowledge on the compounds involved, as well as any bias introduced by examining only one or a few specific scenarios. The Tanimoto index, Dice index, Cosine coefficient and Soergel distance were identified to be the best (and in some sense equivalent) metrics for similarity calculations, i.e. these metrics could produce the rankings closest to the composite (average) ranking of the eight metrics. The similarity metrics derived from Euclidean and Manhattan distances are not recommended on their own, although their variability and diversity from other similarity metrics might be advantageous in certain cases (e.g. for data fusion). Conclusions are also drawn regarding the effects of molecule size, selection method and data pretreatment on the ranking behavior of the studied metrics. Graphical AbstractA visual summary of the comparison of similarity metrics with sum of ranking differences (SRD).

Consistency of QSAR models: Correct split of training and test sets, ranking of models and performance parameters

Recent implementations of QSAR modelling software provide the user with numerous models and a wealth of information. In this work, we provide some guidance on how one should interpret the results of QSAR modelling, compare and assess the resulting models, and select the best and most consistent ones. Two QSAR datasets are applied as case studies for the comparison of model performance parameters and model selection methods. We demonstrate the capabilities of sum of ranking differences (SRD) in model selection and ranking, and identify the best performance indicators and models. While the exchange of the original training and (external) test sets does not affect the ranking of performance parameters, it provides improved models in certain cases (despite the lower number of molecules in the training set). Performance parameters for external validation are substantially separated from the other merits in SRD analyses, highlighting their value in data fusion.

Structure-based Virtual Screening Approaches in Kinase-directed Drug Discovery

Protein kinases are one of the most targeted protein families in current drug discovery pipelines. They are implicated in many oncological, inflammatory, CNS-related and other clinical indications. Virtual screening is a computational technique with a diverse set of available tools that has been shown many times to provide novel starting points for kinase-directed drug discovery. This review starts with a concise overview of the function, structural features and inhibitory mechanisms of protein kinases. In addition to briefly reviewing the practical aspects of structure-based virtual screenings, we discuss several case studies to illustrate the state of the art in the virtual screening for type I, type II, allosteric (type III-V) and covalent (type VI) kinase inhibitors. With this review, we strive to provide a summary of the latest advances in the structure-based discovery of novel kinase inhibitors, as well as a practical tool to anyone who wishes to embark on such an endeavor.

Discovery of Subtype Selective Janus Kinase (JAK) Inhibitors by Structure-Based Virtual Screening

Janus kinase inhibitors represent a promising opportunity for the pharmaceutical intervention of various inflammatory and oncological indications. Subtype selective inhibition of these enzymes, however, is still a very challenging goal. In this study, a novel, customized virtual screening protocol was developed with the intention of providing an efficient tool for the discovery of subtype selective JAK2 inhibitors. The screening protocol involves protein ensemble-based docking calculations combined with an Interaction Fingerprint (IFP) based scoring scheme for estimating ligand affinities and selectivities, respectively. The methodology was validated in retrospective studies and was applied prospectively to screen a large database of commercially available compounds. Six compounds were identified and confirmed in vitro, with an indazole-based hit exhibiting promising selectivity for JAK2 vs JAK1. Having demonstrated that the described methodology is capable of identifying subtype selective chemical starting points with a favorable hit rate (11%), we believe that the presented screening concept can be useful for other kinase targets with challenging selectivity profiles.

Pharmacologic inhibition of STAT5 in acute myeloid leukemia

The transcription factor STAT5 is an essential downstream mediator of many tyrosine kinases (TKs), particularly in hematopoietic cancers. STAT5 is activated by FLT3-ITD, which is a constitutively active TK driving the pathogenesis of acute myeloid leukemia (AML). Since STAT5 is a critical mediator of diverse malignant properties of AML cells, direct targeting of STAT5 is of significant clinical value. Here, we describe the development and preclinical evaluation of a novel, potent STAT5 SH2 domain inhibitor, AC-4–130, which can efficiently block pathological levels of STAT5 activity in AML. AC-4–130 directly binds to STAT5 and disrupts STAT5 activation, dimerization, nuclear translocation, and STAT5-dependent gene transcription. Notably, AC-4–130 substantially impaired the proliferation and clonogenic growth of human AML cell lines and primary FLT3-ITD+ AML patient cells in vitro and in vivo. Furthermore, AC-4–130 synergistically increased the cytotoxicity of the JAK1/2 inhibitor Ruxolitinib and the p300/pCAF inhibitor Garcinol. Overall, the synergistic effects of AC-4–130 with TK inhibitors (TKIs) as well as emerging treatment strategies provide new therapeutic opportunities for leukemia and potentially other cancers.

Validation of tautomeric and protomeric binding modes by free energy calculations. A case study for the structure based optimization of d-amino acid oxidase inhibitors

Optimization of fragment size d-amino acid oxidase (DAAO) inhibitors was investigated using a combination of computational and experimental methods. Retrospective free energy perturbation (FEP) calculations were performed for benzo[d]isoxazole derivatives, a series of known inhibitors with two potential binding modes derived from X-ray structures of other DAAO inhibitors. The good agreement between experimental and computed binding free energies in only one of the hypothesized binding modes strongly support this bioactive conformation. Then, a series of 1-H-indazol-3-ol derivatives formerly not described as DAAO inhibitors was investigated. Binding geometries could be reliably identified by structural similarity to benzo[d]isoxazole and other well characterized series and FEP calculations were performed for several tautomers of the deprotonated and protonated compounds since all these forms are potentially present owing to the experimental pKa values of representative compounds in the series. Deprotonated compounds are proposed to be the most important bound species owing to the significantly better agreement between their calculated and measured affinities compared to the protonated forms. FEP calculations were also used for the prediction of the affinities of compounds not previously tested as DAAO inhibitors and for a comparative structure–activity relationship study of the benzo[d]isoxazole and indazole series. Selected indazole derivatives were synthesized and their measured binding affinity towards DAAO was in good agreement with FEP predictions.Graphical Abstract

Which Performance Parameters Are Best Suited to Assess the Predictive Ability of Models

We have revisited the vivid discussion in the QSAR-related literature concerning the use of external versus cross-validation, and have presented a thorough statistical comparison of model performance parameters with the recently published SRD (sum of (absolute) ranking differences) method and analysis of variance (ANOVA). Two case studies were investigated, one of which has exclusively used external performance merits. The SRD methodology coupled with ANOVA shows unambiguously for both case studies that the performance merits are significantly different, independently from data preprocessing. While external merits are generally less consistent (farther from the reference) than training and cross-validation based merits, a clear ordering and a grouping pattern of them could be acquired. The results presented here corroborate our earlier, recently published findings (SAR QSAR Environ. Res., 2015, 26, 683–700) that external validation is not necessarily a wise choice, and is frequently comparable to a random evaluation of the models.

Property-based characterization of kinase-like ligand space for library design and virtual screening

A property-based desirability scoring scheme has been developed for kinase-focused library design and ligand-based pre-screening of large compound sets. The property distributions of known kinase inhibitors from the ChEMBL Kinase Sarfari database were investigated and used for a desirability function-based score. The scoring scheme is easily interpretable as it accounts for six molecular properties: topological polar surface area and the number of rotatable bonds, hydrogen bond donors, aromatic rings, nitrogen atoms and oxygen atoms. The performance of the Kinase Desirability Score (KiDS) is evaluated on both public and proprietary experimental screening data.

Is soft independent modeling of class analogies a reasonable choice for supervised pattern recognition

A thorough survey of classification data sets and a rigorous comparison of classification methods clearly show the unambiguous superiority of other techniques over soft independent modeling of class analogies (SIMCA) in the case of classification – which is a frequent area of usage for SIMCA, even though it is a class modeling (one class or disjoint class modeling technique). Two non-parametric methods, sum of ranking differences (SRD) and the generalized pairwise correlation method (GPCM) have been used to rank and group the classifiers obtained from six case studies. Both techniques need a supervisor (a reference) and their results support and validate each other, despite being based on entirely different principles and calculation procedures. To eliminate the effect of the chosen reference, comparisons with one variable (classifier) at a time were calculated and presented as heatmaps. Six case studies show unambiguously that SIMCA is inferior to other classification techniques such as linear and quadratic discriminant analyses, multivariate range modeling, etc. This analysis is similar to meta-analyses frequently applied in medical science nowadays; with the notable difference that we did not (and should not) make any distributional assumptions. A well-founded conclusion can be drawn, as we could not find any circumstances when SIMCA is superior to concurrent techniques. Hence, the question in the title is self-explanatory.

Identification of 8‐Hydroxyquinoline Derivatives Active Against Somatic V658F Mutant JAK1‐Dependent Cells

Janus kinases (JAKs) and their gain‐of‐function mutants have been implicated in a range of oncological, inflammatory, and autoimmune conditions, which has sparked great research interest in the discovery and development of small‐molecule JAK inhibitors. Two molecules are currently marketed as JAK inhibitors, but due to the displayed side effects (owing to their suboptimal selectivities among the various JAK subtypes) new JAK inhibitors are still sought after. We present the results of an extensive virtual screening campaign based on a multi‐step screening protocol involving ligand docking. The screening yielded five new, experimentally validated inhibitors of JAK1 with 8‐hydroxyquinoline as a novel hinge‐binding scaffold. The compounds did not only display favorable potencies in a JAK1V658F‐driven cell‐based assay but were also shown to be non‐cytotoxic on rat liver cells.