A data science roadmap for open science organizations engaged in early-stage drug discovery.

The Structural Genomics Consortium is an international open science research organization with a focus on accelerating early-stage drug discovery, namely hit discovery and optimization. We, as many others, believe that artificial intelligence (AI) is poised to be a main accelerator in the field. The question is then how to best benefit from recent advances in AI and how to generate, format and disseminate data to enable future breakthroughs in AI-guided drug discovery. We present here the recommendations of a working group composed of experts from both the public and private sectors. Robust data management requires precise ontologies and standardized vocabulary while a centralized database architecture across laboratories facilitates data integration into high-value datasets. Lab automation and opening electronic lab notebooks to data mining push the boundaries of data sharing and data modeling. Important considerations for building robust machine-learning models include transparent and reproducible data processing, choosing the most relevant data representation, defining the right training and test sets, and estimating prediction uncertainty. Beyond data-sharing, cloud-based computing can be harnessed to build and disseminate machine-learning models. Important vectors of acceleration for hit and chemical probe discovery will be (1) the real-time integration of experimental data generation and modeling workflows within design-make-test-analyze (DMTA) cycles openly, and at scale and (2) the adoption of a mindset where data scientists and experimentalists work as a unified team, and where data science is incorporated into the experimental design.

Optimization of a Novel DEL Hit That Binds in the Cbl-b SH2 Domain and Blocks Substrate Binding.

We were attracted to the therapeutic potential of inhibiting Casitas B-lineage lymphoma proto-oncogene-b (Cbl-b), a RING E3 ligase that plays a critical role in regulating the activation of T cells. However, given that only protein-protein interactions were involved, it was unclear whether inhibition by a small molecule would be a viable approach. After screening an ∼6 billion member DNA-encoded library (DEL) using activated Cbl-b, we identified compound 1 as a hit for which the cis-isomer (2) was confirmed by biochemical and surface plasmon resonance (SPR) assays. Our hit optimization effort was greatly accelerated when we obtained a cocrystal structure of 2 with Cbl-b, which demonstrated induced binding at the substrate binding site, namely, the Src homology-2 (SH2) domain. This was quite noteworthy given that there are few reports of small molecule inhibitors that bind to SH2 domains and block protein-protein interactions. Structure- and property-guided optimization led to compound 27, which demonstrated measurable cell activity, albeit only at high concentrations.

A Multifaceted Hit-Finding Approach Reveals Novel LC3 Family Ligands.

Autophagy-related proteins (Atgs) drive the lysosome-mediated degradation pathway, autophagy, to enable the clearance of dysfunctional cellular components and maintain homeostasis. In humans, this process is driven by the mammalian Atg8 (mAtg8) family of proteins comprising the LC3 and GABARAP subfamilies. The mAtg8 proteins play essential roles in the formation and maturation of autophagosomes and the capture of specific cargo through binding to the conserved LC3-interacting region (LIR) sequence within target proteins. Modulation of interactions of mAtg8 with its target proteins via small-molecule ligands would enable further interrogation of their function. Here we describe unbiased fragment and DNA-encoded library (DEL) screening approaches for discovering LC3 small-molecule ligands. Both strategies resulted in compounds that bind to LC3, with the fragment hits favoring a conserved hydrophobic pocket in mATG8 proteins, as detailed by LC3A-fragment complex crystal structures. Our findings demonstrate that the malleable LIR-binding surface can be readily targeted by fragments; however, rational design of additional interactions to drive increased affinity proved challenging. DEL libraries, which combine small, fragment-like building blocks into larger scaffolds, yielded higher-affinity binders and revealed an unexpected potential for reversible, covalent ligands. Moreover, DEL hits identified possible vectors for synthesizing fluorescent probes or bivalent molecules for engineering autophagic degradation of specific targets.

Regions of conformational flexibility in the proprotein convertase PCSK9 and design of antagonists for LDL cholesterol lowering.

The proprotein convertase subtilisin/kexin type 9 (PCSK9) regulates plasma LDL cholesterol levels by binding to the liver LDL receptor (LDLR) and promoting its degradation. Therefore, PCSK9 has become a compelling new therapeutic target for lipid lowering and the prevention of cardiovascular disease. PCSK9 contains two regions of conformational flexibility, the N-terminal regions of the prodomain and of the catalytic domain. The recognition that the latter region, the so-called P' helix, is able to transition from an α-helical to a disordered state gave rise to new strategies to develop small molecule inhibitors of PCSK9 for lipid lowering. In the ordered state the P' helix is buried in a groove of the PCSK9 catalytic domain located next to the main LDLR binding site. The transition to a disordered state leaves the groove site vacated and accessible for compounds to antagonize LDLR binding. By use of a groove-directed phage display strategy we were able to identify several groove-binding peptides. Based on structural information of PCSK9-peptide complexes, a minimized groove-binding peptide was generated and utilized as an anchor to extend towards the adjacent main LDLR binding site, either by use of a phage-displayed peptide extension library, or by appending organic moieties to yield organo-peptides. Both strategies led to antagonists with pharmacologic activities in cell-based assays. The intricate bipartite mechanism of the potent organo-peptide inhibitors was revealed by structural studies, showing that the core peptide occupies the N-terminal groove, while the organic moiety interacts with the LDLR binding site to create antagonism. These findings validate the PCSK9 groove as an attractive target site and should inspire the development of a new class of small molecule antagonists of PCSK9.

Structure of the essential inner membrane lipopolysaccharide-PbgA complex.

Lipopolysaccharide (LPS) resides in the outer membrane of Gram-negative bacteria where it is responsible for barrier function1,2. LPS can cause death as a result of septic shock, and its lipid A core is the target of polymyxin antibiotics3,4. Despite the clinical importance of polymyxins and the emergence of multidrug resistant strains5, our understanding of the bacterial factors that regulate LPS biogenesis is incomplete. Here we characterize the inner membrane protein PbgA and report that its depletion attenuates the virulence of Escherichia coli by reducing levels of LPS and outer membrane integrity. In contrast to previous claims that PbgA functions as a cardiolipin transporter6-9, our structural analyses and physiological studies identify a lipid A-binding motif along the periplasmic leaflet of the inner membrane. Synthetic PbgA-derived peptides selectively bind to LPS in vitro and inhibit the growth of diverse Gram-negative bacteria, including polymyxin-resistant strains. Proteomic, genetic and pharmacological experiments uncover a model in which direct periplasmic sensing of LPS by PbgA coordinates the biosynthesis of lipid A by regulating the stability of LpxC, a key cytoplasmic biosynthetic enzyme10-12. In summary, we find that PbgA has an unexpected but essential role in the regulation of LPS biogenesis, presents a new structural basis for the selective recognition of lipids, and provides opportunities for future antibiotic discovery.

Cyanopyrrolidine Inhibitors of Ubiquitin Specific Protease 7 Mediate Desulfhydration of the Active-Site Cysteine.

Ubiquitin specific protease 7 (USP7) regulates the protein stability of key cellular regulators in pathways ranging from apoptosis to neuronal development, making it a promising therapeutic target. Here we used an engineered, activated variant of the USP7 catalytic domain to perform structure-activity studies of electrophilic peptidomimetic inhibitors. Employing this USP7 variant, we found that inhibitors with a cyanopyrrolidine warhead unexpectedly promoted a ß-elimination reaction of the initial covalent adducts, thereby converting the active-site cysteine residue to dehydroalanine. We determined that this phenomenon is specific for the USP7 catalytic cysteine and that structural features of the inhibitor and protein microenvironment impact elimination rates. Using comprehensive docking studies, we propose that the characteristic conformational dynamics of USP7 allow access to conformations that promote the ligand-induced elimination. Unlike in conventional reversible-covalent inhibition, the compounds described here irreversibly destroy a catalytic residue while simultaneously converting the inhibitor to a nonelectrophilic byproduct. Accordingly, this unexpected finding expands the scope of covalent inhibitor modalities and offers intriguing insights into enzyme-inhibitor dynamics.

Design of Organo-Peptides As Bipartite PCSK9 Antagonists.

Proprotein convertase subtilisin/kexin 9 (PCSK9) has become an important therapeutic target for lipid lowering, since it regulates low-density lipoprotein cholesterol (LDL-c) levels by binding to liver LDL receptors (LDLR) and effecting their intracellular degradation. However, the development of small molecule inhibitors is hampered by the lack of attractive PCSK9 target sites. We recently discovered helical peptides that are able to bind to a cryptic groove site on PCSK9, which is situated in proximity to the main LDLR binding site. Here, we designed potent bipartite PCSK9 inhibitors by appending organic moieties to a helical groove-binding peptide to reach a hydrophobic pocket in the proximal LDLR binding region. The ultimately designed 1-amino-4-phenylcyclohexane-1-carbonyl extension improved the peptide affinity by >100-fold, yielding organo-peptide antagonists that potently inhibited PCSK9 binding to LDLR and preserved cellular LDLR. These new bipartite antagonists have reduced mass and improved potency compared to the first-generation peptide antagonists, further validating the PCSK9 groove as a viable therapeutic target site.

Disrupting Gram-Negative Bacterial Outer Membrane Biosynthesis through Inhibition of the Lipopolysaccharide Transporter MsbA.

There is a critical need for new antibacterial strategies to counter the growing problem of antibiotic resistance. In Gram-negative bacteria, the outer membrane (OM) provides a protective barrier against antibiotics and other environmental insults. The outer leaflet of the outer membrane is primarily composed of lipopolysaccharide (LPS). Outer membrane biogenesis presents many potentially compelling drug targets as this pathway is absent in higher eukaryotes. Most proteins involved in LPS biosynthesis and transport are essential; however, few compounds have been identified that inhibit these proteins. The inner membrane ABC transporter MsbA carries out the first essential step in the trafficking of LPS to the outer membrane. We conducted a biochemical screen for inhibitors of MsbA and identified a series of quinoline compounds that kill Escherichia coli through inhibition of its ATPase and transport activity, with no loss of activity against clinical multidrug-resistant strains. Identification of these selective inhibitors indicates that MsbA is a viable target for new antibiotics, and the compounds we identified serve as useful tools to further probe the LPS transport pathway in Gram-negative bacteria.

A Novel Inhibitor of the LolCDE ABC Transporter Essential for Lipoprotein Trafficking in Gram-Negative Bacteria.

The outer membrane is an essential structural component of Gram-negative bacteria that is composed of lipoproteins, lipopolysaccharides, phospholipids, and integral ß-barrel membrane proteins. A dedicated machinery, called the Lol system, ensures proper trafficking of lipoproteins from the inner to the outer membrane. The LolCDE ABC transporter is the inner membrane component, which is essential for bacterial viability. Here, we report a novel pyrrolopyrimidinedione compound, G0507, which was identified in a phenotypic screen for inhibitors of Escherichia coli growth followed by selection of compounds that induced the extracytoplasmic σE stress response. Mutations in lolC, lolD, and lolE conferred resistance to G0507, suggesting LolCDE as its molecular target. Treatment of E. coli cells with G0507 resulted in accumulation of fully processed Lpp, an outer membrane lipoprotein, in the inner membrane. Using purified protein complexes, we found that G0507 binds to LolCDE and stimulates its ATPase activity. G0507 still binds to LolCDE harboring a Q258K substitution in LolC (LolCQ258K), which confers high-level resistance to G0507 in vivo but no longer stimulates ATPase activity. Our work demonstrates that G0507 has significant promise as a chemical probe to dissect lipoprotein trafficking in Gram-negative bacteria.

Selectively Modulating Conformational States of USP7 Catalytic Domain for Activation.

Ubiquitin-specific protease 7 (USP7) deubiquitinase activity is controlled by a number of regulatory factors, including stimulation by intramolecular accessory domains. Alone, the USP7 catalytic domain (USP7cd) shows limited activity and apo USP7cd crystal structures reveal a disrupted catalytic triad. By contrast, ubiquitin-conjugated USP7cd structures demonstrate the canonical cysteine protease active-site geometry; however, the structural features of the USP7cd that stabilize the inactive conformation and the mechanism of transition between inactive and active states remain unclear. Here we use comparative structural analyses, molecular dynamics simulations, and in silico sequence re-engineering via directed sampling by RosettaDesign to identify key molecular determinants of USP7cd activation and successfully engineer USP7cd for improved activity. Full kinetic analysis and multiple X-ray crystal structures of our designs indicate that electrostatic interactions in the distal "switching loop" region and local packing in the hydrophobic core mediate subtle but significant conformational changes that modulate USP7cd activation.

Selective Homogeneous Assay for Circulating Endopeptidase Fibroblast Activation Protein (FAP).

Fibroblast Activation Protein (FAP) is a membrane-bound serine protease whose expression is often elevated in activated fibroblasts associated with tissue remodeling in various common diseases such as cancer, arthritis and fibrosis. Like the closely related dipeptidyl peptidase DPPIV, the extracellular domain of FAP can be released into circulation as a functional enzyme, and limited studies suggest that the circulating level of FAP correlates with the degree of tissue fibrosis. Here we describe a novel homogeneous fluorescence intensity assay for circulating FAP activity based on a recently identified natural substrate, FGF21. This assay is unique in that it can effectively distinguish endopeptidase activity of FAP from that of other related enzymes such as prolyl endopeptidase (PREP) and was validated using Fap-deficient mice. Structural modeling was used to elucidate the mechanistic basis for the observed specificity in substrate recognition by FAP, but not by DPPIV or PREP. Finally, the assay was used to detect elevated FAP activity in human patients diagnosed with liver cirrhosis and to determine the effectiveness of a chemical inhibitor for FAP in mice. We propose that the assay presented here could thus be utilized for diagnosis of FAP-related pathologies and for the therapeutic development of FAP inhibitors.

Discovery of a cryptic peptide-binding site on PCSK9 and design of antagonists.

Proprotein convertase subtilisin/kexin type 9 (PCSK9) regulates plasma LDL cholesterol (LDL-c) levels by promoting the degradation of liver LDL receptors (LDLRs). Antibodies that inhibit PCSK9 binding to the EGF(A) domain of the LDLR are effective in lowering LDL-c. However, the discovery of small-molecule therapeutics is hampered by difficulty in targeting the relatively flat EGF(A)-binding site on PCSK9. Here we demonstrate that it is possible to target this site, based on the finding that the PCSK9 P' helix displays conformational flexibility. As a consequence, the vacated N-terminal groove of PCSK9, which is adjacent to the EGF(A)-binding site, is in fact accessible to small peptides. In phage-display experiments, the EGF(A)-mimicking peptide Pep2-8 was used as an anchor peptide for the attachment of an extension peptide library directed toward the groove site. Guided by structural information, we further engineered the identified groove-binding peptides into antagonists, which encroach on the EGF(A)-binding site and inhibit LDLR binding.

Minimizing CYP2C9 Inhibition of Exposed-Pyridine NAMPT (Nicotinamide Phosphoribosyltransferase) Inhibitors.

NAMPT inhibitors may show potential as therapeutics for oncology. Throughout our NAMPT inhibitor program, we found that exposed pyridines or related heterocyclic systems in the left-hand portion of the inhibitors are necessary pharmacophores for potent cellular NAMPT inhibition. However, when combined with a benzyl group in the center of the inhibitors, such pyridine-like moieties also led to consistent and potent inhibition of CYP2C9. In an attempt to reduce CYP2C9 inhibition, a parallel synthesis approach was used to identify central benzyl group replacements with increased Fsp3. A spirocyclic central motif was thus discovered that was combined with left-hand pyridines (or pyridine-like systems) to provide cellularly potent NAMPT inhibitors with minimal CYP2C9 inhibition. Further optimization of potency and ADME properties led to the discovery of compound 68, a highly potent NAMPT inhibitor with outstanding efficacy in a mouse tumor xenograft model and lacking measurable CYP2C9 inhibition at the concentrations tested.

Activation Mechanism of Oncogenic Deletion Mutations in BRAF, EGFR, and HER2.

Activating mutations in protein kinases drive many cancers. While how recurring point mutations affect kinase activity has been described, the effect of in-frame deletions is not well understood. We show that oncogenic deletions within the ß3-αC loop of HER2 and BRAF are analogous to the recurrent EGFR exon 19 deletions. We identify pancreatic carcinomas with BRAF deletions mutually exclusive with KRAS mutations. Crystal structures of BRAF deletions reveal the truncated loop restrains αC in an active "in" conformation, imparting resistance to inhibitors like vemurafenib that bind the αC "out" conformation. Characterization of loop length explains the prevalence of five amino acid deletions in BRAF, EGFR, and HER2 and highlights the importance of this region for kinase activity and inhibitor efficacy.

Identification of nicotinamide phosphoribosyltransferase (NAMPT) inhibitors with no evidence of CYP3A4 time-dependent inhibition and improved aqueous solubility.

Herein we report the optimization efforts to ameliorate the potent CYP3A4 time-dependent inhibition (TDI) and low aqueous solubility exhibited by a previously identified lead compound from our NAMPT inhibitor program (1, GNE-617). Metabolite identification studies pinpointed the imidazopyridine moiety present in 1 as the likely source of the TDI signal, and replacement with other bicyclic systems was found to reduce or eliminate the TDI finding. A strategy of reducing the number of aromatic rings and/or lowering cLogD7.4 was then employed to significantly improve aqueous solubility. These efforts culminated in the discovery of 42, a compound with no evidence of TDI, improved aqueous solubility, and robust efficacy in tumor xenograft studies.

Small-Molecule Library Subset Screening as an Aid for Accelerating Lead Identification.

Several small-compound library subsets (14,000 to 56,000) have been established to complement screening of a larger Genentech corporate library (~1,300,000). Two validation sets (~1% of the total library) containing compounds representative of the main library were chosen by selection of plates or individual compounds. Use of these subsets guided selection of assay configuration, validated assay reproducibility, and provided estimates of hit rates expected from our full library. A larger diversity subset representing the scaffold diversity of the full library (3.4% of the total) was designed for screening more challenging targets with limited reagent availability or low-throughput assays. Retrospective analysis of this subset showed hit rates similar to those of the main library while recovering a higher proportion of hit scaffolds. Finally, a property-restricted diversity set called the "in-between library" was established to identify ligand-efficient compounds of molecular size between those typically found in fragment and high-throughput screening libraries. It was screened at fivefold higher concentrations than the main library to facilitate identification of less potent yet ligand-efficient compounds. Taken together, this work underscores the value of generating multiple purpose-focused, diversity-based library subsets that are designed using computational approaches coupled with internal screening data analyses to accelerate the lead discovery process.

Fragment-based design of 3-aminopyridine-derived amides as potent inhibitors of human nicotinamide phosphoribosyltransferase (NAMPT).

The fragment-based identification of two novel and potent biochemical inhibitors of the nicotinamide phosphoribosyltransferase (NAMPT) enzyme is described. These compounds (51 and 63) incorporate an amide moiety derived from 3-aminopyridine, and are thus structurally distinct from other known anti-NAMPT agents. Each exhibits potent inhibition of NAMPT biochemical activity (IC50=19 and 15 nM, respectively) as well as robust antiproliferative properties in A2780 cell culture experiments (IC50=121 and 99 nM, respectively). However, additional biological studies indicate that only inhibitor 51 exerts its A2780 cell culture effects via a NAMPT-mediated mechanism. The crystal structures of both 51 and 63 in complex with NAMPT are also independently described.

Fragment-based identification of amides derived from trans-2-(pyridin-3-yl)cyclopropanecarboxylic acid as potent inhibitors of human nicotinamide phosphoribosyltransferase (NAMPT).

Potent, trans-2-(pyridin-3-yl)cyclopropanecarboxamide-containing inhibitors of the human nicotinamide phosphoribosyltransferase (NAMPT) enzyme were identified using fragment-based screening and structure-based design techniques. Multiple crystal structures were obtained of initial fragment leads, and this structural information was utilized to improve the biochemical and cell-based potency of the associated molecules. Many of the optimized compounds exhibited nanomolar antiproliferative activities against human tumor lines in in vitro cell culture experiments. In a key example, a fragment lead (13, KD = 51 µM) was elaborated into a potent NAMPT inhibitor (39, NAMPT IC50 = 0.0051 µM, A2780 cell culture IC50 = 0.000 49 µM) which demonstrated encouraging in vivo efficacy in an HT-1080 mouse xenograft tumor model.

Identification of a small peptide that inhibits PCSK9 protein binding to the low density lipoprotein receptor.

PCSK9 (proprotein convertase subtilisin/kexin type 9) is a negative regulator of the hepatic LDL receptor, and clinical studies with PCSK9-inhibiting antibodies have demonstrated strong LDL-c-lowering effects. Here we screened phage-displayed peptide libraries and identified the 13-amino acid linear peptide Pep2-8 as the smallest PCSK9 inhibitor with a clearly defined mechanism of inhibition that has been described. Pep2-8 bound to PCSK9 with a KD of 0.7 µm but did not bind to other proprotein convertases. It fully restored LDL receptor surface levels and LDL particle uptake in PCSK9-treated HepG2 cells. The crystal structure of Pep2-8 bound to C-terminally truncated PCSK9 at 1.85 Å resolution showed that the peptide adopted a strand-turn-helix conformation, which is remarkably similar to its solution structure determined by NMR. Consistent with the functional binding site identified by an Ala scan of PCSK9, the structural Pep2-8 contact region of about 400 Å(2) largely overlapped with that contacted by the EGF(A) domain of the LDL receptor, suggesting a competitive inhibition mechanism. Consistent with this, Pep2-8 inhibited LDL receptor and EGF(A) domain binding to PCSK9 with IC50 values of 0.8 and 0.4 µm, respectively. Remarkably, Pep2-8 mimicked secondary structural elements of the EGF(A) domain that interact with PCSK9, notably the ß-strand and a discontinuous short α-helix, and it engaged in the same ß-sheet hydrogen bonds as EGF(A) does. Although Pep2-8 itself may not be amenable to therapeutic applications, this study demonstrates the feasibility of developing peptidic inhibitors to functionally relevant sites on PCSK9.

Identification of amides derived from 1H-pyrazolo[3,4-b]pyridine-5-carboxylic acid as potent inhibitors of human nicotinamide phosphoribosyltransferase (NAMPT).

Potent, 1H-pyrazolo[3,4-b]pyridine-containing inhibitors of the human nicotinamide phosphoribosyltransferase (NAMPT) enzyme were identified using structure-based design techniques. Many of these compounds exhibited nanomolar antiproliferation activities against human tumor lines in in vitro cell culture experiments, and a representative example (compound 26) demonstrated encouraging in vivo efficacy in a mouse xenograft tumor model derived from the A2780 cell line. This molecule also exhibited reduced rat retinal exposures relative to a previously studied imidazo-pyridine-containing NAMPT inhibitor. Somewhat surprisingly, compound 26 was only weakly active in vitro against mouse and monkey tumor cell lines even though it was a potent inhibitor of NAMPT enzymes derived from these species. The compound also exhibited only minimal effects on in vivo NAD levels in mice, and these changes were considerably less profound than those produced by an imidazo-pyridine-containing NAMPT inhibitor. The crystal structures of compound 26 and the corresponding PRPP-derived ribose adduct in complex with NAMPT were also obtained.