Novel inhibitors of an inflammatory kinase – found first try

p38α MAP kinase (MAPK14) is a central regulator of pro-inflammatory cytokines (including TNF-α, IL-1, and IL-6) and attracted decades of drug discovery investment for inflammatory diseases. Every clinical program ultimately failed, not for lack of potent inhibitors, but for reasons rooted in the target’s biology: compensatory upregulation of parallel MAPK pathways causes cytokine levels to rebound in hepatocytes and cardiomyocytes produces mechanism-based toxicity that more selective compounds cannot escape.

p38α therefore remains a valuable benchmark target being well-characterized, with abundant structural data and robust assays even as its therapeutic utility as narrowed. What the field lacked was not better inhibitors per se, but genuinely novel chemical matter to probe its biology from new angles. Existing series shared common hinge-binding motifs, a naïve, unbiased screen offered the prospect of something structurally distinct.

Vipergen screened p38α against Lib022, a yoctoReactor-generated DNA-encoded library (DEL) comprising 12.6 million compounds, each uniquely identified by its covalently attached DNA barcode. Crucially, the library was not designed with kinases in mind: no prior structural knowledge of the target was incorporated into its design, making this a genuine test of the platform’s ability to discover novel chemical matter from scratch.

Screening was performed using Binder Trap Enrichment (BTE), Vipergen’s proprietary homogeneous solution-phase selection technology. Rather than immobilizing the target to a solid support, which risks denaturation, matrix binding artefacts, and low signal-to-noise, BTE traps target–ligand complexes within individual water-in-oil emulsion droplets.

Of 24 resynthesized compounds, 22 were confirmed as p38α inhibitors yielding a hit confirmation rate of 92%, underlying the low false-positive rate inherent to BTE. The primary hit series yielded inhibitors spanning from low micromolar to single-digit nanomolar potency in biochemical assays, with lead compound VPC00628 achieving a cellular IC50 of 7 nM in a TNF-α secretion assay in human THP-1 monocytic cells.

Critically, the structural novelty of the hits went beyond selectivity numbers. X-ray crystallography of the p38α-VPC00628 complex at 1.5 Å resolution revealed a canonical type-II (‘DFG-out’) binding mode, yet one achieved through a strikingly atypical set of interactions. The compounds lack the hinge-binding motifs considered hallmarks of kinase inhibitors, instead forming a hydrogen bond-network via its pyrazole nitrogen with the backbone M109, while a cyclohexane decoration occupies the hydrophobic pocket opened by the DFG-out conformation.

Most remarkably, VPC00628 induced a pronounced distortion of the P-loop, a structural alteration not previously observed in complexes with established type-II inhibitors such as BIRB796 or sorafenib. This generated a channel that, in the original DNA-encoded compound, accommodated the linker to the DNA tag itself. This is a clear illustration of how DEL-derived hits can encode structural information about the screening format directly into their binding geometry.

This study is perhaps the clearest demonstration of what BTE can deliver when given an unconstrained library and a well-defined target. No kinase-biased building blocks were incorporated. No kinase-specific pharmacophores were used as a starting point. The hit series emerged from chemistry selected on binding affinity alone, and yet yielded compounds potent enough for cellular validation, all without a single round of medicinal chemistry optimization.

The modular structure of the yoctoReactor library also means that follow-up chemistry is immediately tractable: each building block position can be varied independently, providing a direct path from screening hit to optimized lead.