First fully-automated AI/ML virtual screening cascade implemented at a drug discovery centre in Africa.

Streamlined data-driven drug discovery remains challenging, especially in resource-limited settings. We present ZairaChem, an artificial intelligence (AI)- and machine learning (ML)-based tool for quantitative structure-activity/property relationship (QSAR/QSPR) modelling. ZairaChem is fully automated, requires low computational resources and works across a broad spectrum of datasets. We describe an end-to-end implementation at the H3D Centre, the leading integrated drug discovery unit in Africa, at which no prior AI/ML capabilities were available. By leveraging in-house data collected over a decade, we have developed a virtual screening cascade for malaria and tuberculosis drug discovery comprising 15 models for key decision-making assays ranging from whole-cell phenotypic screening and cytotoxicity to aqueous solubility, permeability, microsomal metabolic stability, cytochrome inhibition, and cardiotoxicity. We show how computational profiling of compounds, prior to synthesis and testing, can inform progression of frontrunner compounds at H3D. This project is a first-of-its-kind deployment at scale of AI/ML tools in a research centre operating in a low-resource setting.

Molecular Modeling of the Shigella flexneri Serogroup 3 and 5 O-Antigens and Conformational Relationships for a Vaccine Containing Serotypes 2a and 3a.

The pathogenic bacterium Shigella flexneri is a leading global cause of diarrheal disease. The O-antigen is the primary vaccine target and distinguishes the 30 serotypes reported. Except for serotype 6, all S. flexneri serotypes have a common backbone repeating unit (serotype Y), with variations in substitution creating the various serotypes. A quadrivalent vaccine containing serotypes 2a and 3a (as well as 6 and Shigella sonnei) is proposed to provide broad protection against non-vaccine S. flexneri serotypes through shared epitopes and conformations. Here we model the O-antigen (O-Ag) conformations of serogroups 3 and 5: a continuation of our ongoing systematic study of the S. flexneri O-antigens that began with serogroup 2. Our simulations show that S. flexneri serogroups 2, 3, and 5 all have flexible O-Ags, with substitutions of the backbone altering the chain conformations in different ways. Our analysis suggests three general heuristics for the effects of substitution on the Shigella O-Ag conformations: (1) substitution on rhamnose C reduces the extension of the O-Ag chain; (2) substitution at O-3 of rhamnose A restricts the O-Ags to predominantly helical conformations, (3) substitution at O-3 of rhamnose B has only a slight effect on conformation. The common O-Ag conformations across serotypes identified in this work support the assumption that a quadrivalent vaccine containing serotypes 2a and 3a could provide coverage against S. flexneri serotype 3b and serogroup 5.

Effects of Glucosylation and O-Acetylation on the Conformation of Shigella flexneri Serogroup 2 O-Antigen Vaccine Targets.

Shigellosis is an enteric disease with high morbidity and mortality, particularly in developing countries. There is currently no licensed vaccine available. Most infection is caused by Shigella flexneri, of which 30 serotypes have been recognized based on O-antigen polysaccharide structure. Almost all S. flexneri serotypes share the same repeating unit backbone (serotype Y), with varying glucosylation, O-acetylation and phosphorylation. The O-antigen is the primary vaccine target; the vaccine valency (and hence cost) can be reduced by cross-protection. Our planned systematic conformational study of S. flexneri starts here with 2a, the dominant cause of infection globally. We employ microsecond molecular dynamics simulations to compare the conformation of the unsubstituted serotype Y backbone with the serogroup 2 O-antigens, to investigate the effect of glucosylation and O-acetylation (O-factor 9) on conformation. We find that serotype Y is highly flexible, whereas glucosylation in 2a restricts flexibility and induces C-curve conformations. Further, the glucose side-chains adopt two distinct conformations, corroborated by the antibody-bound crystal structure data. Additional substitution on O-3 of rhamnose A (whether O-acetylation in 2a or glucosylation in 2b) induces helical conformations. Our results suggest that the O-3-acetylated 2a antigen will elicit cross-protection against 2b, as well as other serotypes containing O-factor 9.

Modeling the conformations of Neisseria meningitidis serogroup a CPS and a carba-analogue: Implications for vaccine development.

Neisseria meningitidis is a major cause of bacterial meningitis worldwide, especially in Africa. The capsular polysaccharide is the main virulence factor and the target antigen for polysaccharide- and conjugate vaccines. Three tetravalent conjugate vaccines against serogroups A, C, Y and W have been licensed and the monovalent MenAfriVac® was introduced to address the high burden of serogroup A disease in the Meningitis Belt of sub-Saharan Africa. Three of these four vaccines are lyophilized due to the instability of the serogroup A antigen (MenA) in aqueous solution, resulting in a two vial presentation with concomitant additional costs for storage and distribution. Replacement of the saccharide ring oxygen with a methylene group is a promising approach to preparing a stable oligosaccharide MenA analogue (Carba-MenA) vaccine suitable for a liquid formulation. However, to be effective, Carba-MenA must elicit an immune response that is cross-reactive to the native MenA. Here we employ microsecond molecular dynamics simulations of ten repeats of MenA and Carba-MenA to establish that there are significant differences in the conformation and dynamics of these antigens in solution. Carba-MenA has a more random extended, conformation than MenA; MenA has a significant population of compact S-bend conformations that are absent in the analogue. We also find that the disaccharides are poor models of the conformational behaviour of longer chains. This information is relevant for the rational design of optimal analogues for conjugate vaccines.

Conformations of Neisseria meningitidis serogroup A and X polysaccharides: The effects of chain length and O-acetylation.

Neisseria meningitidis is a major cause of bacterial meningitis worldwide especially in Africa. The capsular polysaccharide (CPS) is the main virulence factor and the target antigen for polysaccharide and conjugate vaccines. The high burden of serogroup A disease in the Meningitis Belt of sub-Saharan Africa led to the introduction of MenAfriVac®, which has successfully reduced the number of cases of group A disease. However, several outbreaks caused by other serogroups have been reported, including those due to serogroup X. The capsular polysaccharides of serogroups A and X are both homopolymers of amino sugars (α-D-ManNAc and α-D-GlcNAc) containing phosphodiester linkages at C-6 and C-4, respectively. The similarity of the primary structures of the two polysaccharides suggests that serogroup A vaccination may provide cross-protection against serogroup X disease. Molecular dynamics simulations of a series of serogroup A and X oligosaccharides reveal that the MenA CPS behaves as a flexible random coil which becomes less conformationally defined as the length increases, whereas serogroup X forms a more stable regular helical structure. The presence of the MenX helix is supported by NMR analysis; it has four residues per turn and becomes more stable as the chain length increases. Licensed MenA vaccines are largely O-acetylated at C-3: simulations show that these O-acetyl groups are highly solvent exposed and their presence favors more extended conformations compared to the more compact conformations of MenA without O-acetylation. These findings may have implications for the design of optimal conjugate vaccines.

Genetic and structural elucidation of capsular polysaccharides from Streptococcus pneumoniae serotype 23A and 23B, and comparison to serotype 23F.

Streptococcus pneumoniae is a globally important encapsulated human pathogen with approximately 100 different serotypes recognized. Serogroup 23 consists of serotype 23F, present in licensed vaccines, and emerging serotypes 23A and 23B. Here, we report the previously unknown structures of the pneumococcal capsular polysaccharides serotype 23A and 23B determined using genetic analysis, NMR spectroscopy, composition and linkage analysis and Smith degradation (of polysaccharide 23A). The structure of the serotype 23A capsular polysaccharide is: →4)-ß-D-Glcp-(1→3)-[[α-L-Rhap-(1→2)]-[Gro-(2→P→3)]-ß-D-Galp-(1→4)]-ß-L-Rhap-(1→. This structure differs from polysaccharide 23F as it features a disaccharide backbone and the di-substituted ß-Gal is linked to ß-Rha as a side chain. This is due to the different polymerization position catalysed by the unusually divergent repeat unit polymerase Wzy in the 23A cps biosynthesis locus. Steric crowding in 23A, confirmed by molecular models, causes the NMR signal for H-1 of the di-substituted 2,3-ß-Gal to resonate in the α-anomeric region. The structure of the serotype 23B capsular polysaccharide is the same as 23F, but without the terminal α-Rha: →4)-ß-D-Glcp-(1→4)-[Gro-(2→P→3)]-ß-D-Galp-(1→4)-ß-L-Rhap-(1→. The immunodominant terminal α-Rha of 23F is more sterically crowded in 23A and absent in 23B. This may explain the reported typing cross reactions for serotype 23F: slight with 23A and none with 23B.