From Genome to Drugs: New Approaches in Antimicrobial Discovery

Abstract

Decades of successful use of antibiotics is currently challenged by the emergence of increasingly resistant bacterial strains. Novel drugs are urgently required but, in a scenario where private investment in the development of new antimicrobials is declining, efforts to combat drug-resistant infections become a worldwide public health problem. Reasons behind unsuccessful new antimicrobial development projects range from inadequate selection of the molecular targets to a lack of innovation. In this context, increasingly available omics data for multiple pathogens has created new drug discovery and development opportunities to fight infectious diseases. Identification of an appropriate molecular target is currently accepted as a critical step of the drug discovery process. Here, we review how diverse layers of multi-omics data in conjunction with structural/functional analysis and systems biology can be used to prioritize the best candidate proteins. Once the target is selected, virtual screening can be used as a robust methodology to explore molecular scaffolds that could act as inhibitors, guiding the development of new drug lead compounds. This review focuses on how the advent of omics and the development and application of bioinformatics strategies conduct a "big-data era" that improves target selection and lead compound identification in a cost-effective and shortened timeline.

Keywords: drug discovery; drug target; metabolic reconstruction; structural modeling; target prioritization; virtual screening.

FIGURE 1

A general sketch of Target-Pathogen integrated with LigQ pipeline. Structural druggability and metabolic analyses are integrated with available experimental data and in silico analysis data. After all, data is integrated into Target-Pathogen, a user-designed scoring function is used to weight different features to obtain a ranked list of candidate drug targets. Once the target is selected, LigQ pipeline finds all known binders of similar proteins in the PDB and ChEMBL. Target and putative binders can be used in further molecular docking assays.

FIGURE 2

Potential druggable targets and putative lead compounds to combat Bartonella bacilliformis. (A) Dihydrofolate reductase (FolA) structure. Crystallographic structure and a set of attractive features for target prioritization are shown. (B) Number of proteins in Bartonella bacilliformis genome with desirable properties for drug targets. Different filters are sequentially applied to obtain a shortlist of druggable, essential, and low identities with proteins in the human genome. The last filter is applied to get the list of proteins with putative binders. (C) Seed Compounds for Bartonella bacilliformis. Venn diagram of the number of seed compounds corresponding to different sets. Binders are classified into two seed groups. Seed II are those drugs that bind any protein that harbor the same Pfam domains with Bb proteins and have been co-crystallized with such proteins in the PDB. Seed IV is the set of drugs that bind any protein that shares Pfam domains with Bartonella bacilliformis proteins and was reported as active in Chembl. (D) Number of proteins in the Bartonella bacilliformis genome for which a ligand can be predicted. The top panel corresponds to drugs in Seed II. The bottom panel corresponds to drugs in Seed IV.

FIGURE 3

Potential druggable targets and putative lead compounds to combat Mycobacterium tuberculosis. (A) Inositol-3-phosphate synthase (Ino1) structure. Crystallographic structure and a set of attractive features for target prioritization are shown. (B) Number of proteins in Mycobacterium tuberculosis genome with desirable properties for drug targets. Different filters are sequentially applied to obtain a shortlist of druggable, essential, and low identities with proteins in the human genome. The last filter is applied to get the list of proteins with putative binders (C) Seed Compounds for Mycobacterium tuberculosis. Venn diagram of the number of seed compounds corresponding to different sets. Binders are classified in groups according to the degree of protein similarity [starting from high identity >60% homologs to binders to the same domain in PFAM (Mistry et al., 2020)] and available information (such as the structure of the protein-ligand complex) in different databases such as Protein Data Bank (PDB), Pfam, ChEMBL (EMBL-EBI). Venn diagram of the number of seed compounds corresponding to different sets. Binders are classified into two seed groups. Seed I and III are obtained through the direct search of the protein of interest by its corresponding identifier (ID) for each base and ChEMBL. Seed II are those drugs that bind any protein that harbor the same Pfam domains with Bb proteins and have been co-crystallized with such proteins in the PDB. Seed IV is the set of drugs that bind any protein that shares Pfam domains with Mycobacterium proteins and was reported as active in Chembl. (D) Number of proteins in the Mycobacterium tuberculosis genome for which a ligand can be predicted. The top panel corresponds to drugs in Seed II. The bottom panel corresponds to drugs in Seed IV.