Computational Strategies for Broad Spectrum Venom Phospholipase A2 Inhibitors

Abstract

Snakebite envenoming is a persistent cause of mortality and morbidity worldwide due to the logistical challenges and costs of current antibody-based treatments. Their persistence motivates a broad interest in the discovery of inhibitors against multispecies venom phospholipase A₂ (PLA₂), which are underway as an alternative or supplemental treatment to improve health outcomes. Here, we present new computational strategies for improved inhibitor classification for challenging metalloenzyme targets across many species, including both a new method to utilize existing molecular docking, and subsequent data normalization. These methods were improved to support experimental screening efforts estimating the broader efficacy of candidate PLA₂ inhibitors against diverse viper and elapid venoms.

Figure 1

Rendering of a PLA₂ (RCSB: 1TGM) featuring a phospholipid substrate docked at the active site (a), and depiction of the enzyme active state for phospholipid hydrolysis (b) with aspartate (red), histidine (blue), and calcium ion (green) highlighted.

Figure 2

Structural comparison of common PLA₂ isoforms: (a) renderings of aligned PLA₂ structures from viper (blue) and elapid (red) snake families, and (b) correlation plot of sequence- and structure-based comparison metrics within (red and blue) or between (black) family groups (see Supporting Information Section S3 for methodological details). Error bars represent the interquartile range within the PLA₂ structures.

Figure 3

Distributions of candidate inhibitors by experimental inhibition (a, n_candidates = 192), and protomer charge (b, n_protomers = 961). We note that protomers from acidic polyphenols (e.g., tannic acid), were enumerated as highly charged species from Dimorphite-DL. Further details the prevalence of common drug-likeness parameters in candidate inhibitors are shown in Figure S2.

Figure 4

(a) Receiver operating characteristic (ROC) plot ordering of strong PLA₂ inhibitors with different scoring functions^, (Table S4) and associated area-under-the-curve values (AUC); (b) the S^M–1TGM complex from PLANTS1.2/chemPLP, showing the proton transfer hydrogen bonds (blue dashed lines) and relevant distances between the substrate and active site (red dashed lines).

Figure 5

Visualization of (a) traditional and (b) displacement docking, with the ligand shown in orange (a), the penalty mask as a gray volume (b), and the substrate S^M (docked in the displacement docking step) in pink.

Figure 6

Distribution of scores (with boxes indicating the interquartile range of the data) for strong (purple) and weak (gray) inhibitors from either (a) displacement or (b) traditional direct docking approaches, as listed in Supporting Information Table S6. Their correlation is visualized by (c) a scatterplot with a linear fit (R² = 0.09). Lastly, the performance of the two docking approaches alone or in combination as a ratio (eq 1) is assessed by (d) a ROC plot for the discovery of strong inhibitors against 1TGM, with AUC values indicated.

Figure 7

An example of substrate standardization for D. russelii PLA₂. Individual substrate-based cutoff point-scoring metrics (a), including: the traditional direct docking score (obtained with chemPLP), our displacement docking score (this work), our combined ratio score (eq 1, this work), and steric and metal interaction scores obtained from the traditional direct docking with the chemPLP scoring function (see Tables S6 and S7). In these figures, the substrate-based cutoff is shown as a blue line, with the mean of each data set indicated by a black line. Comparison of the total number of compounds (all candidates) and the counts of candidate inhibitors that pass the substrate-based cutoff (b), for each metric listed (separated for strong and weak inhibitors) and with calculated enrichment (eq 2) indicated at the x-axis labels.

Figure 8

Multispecies screening results: (a) diamond boxplots of points-based scoring (eq 3) of strong (“+”) and weak inhibitors (“–”) against viper (blues) and elapid (reds) PLA₂s (Tables S7–S14), and (b) heatmaps showing the pairwise correlation of points-based scores between inhibitors, as obtained among the PLA₂s considered (referred to by their PDB IDs; Table 1).