Idris Raji

Ionizable lipids are a key component of lipid nanoparticles, the leading nonviral messenger RNA delivery technology. Here, to advance the identification of ionizable lipids beyond current methods, which rely on experimental screening and/or rational design, we introduce lipid optimization using neural networks, a deep-learning strategy for ionizable lipid design. We created a dataset of >9,000 lipid nanoparticle activity measurements and used it to train a directed message-passing neural network for prediction of nucleic acid delivery with diverse lipid structures. Lipid optimization using neural networks predicted RNA delivery in vitro and in vivo and extrapolated to structures divergent from the training set. We evaluated 1.6 million lipids in silico and identified two structures, FO-32 and FO-35, with local mRNA delivery to the mouse muscle and nasal mucosa. FO-32 matched the state of the art for nebulized mRNA delivery to the mouse lung, and both FO-32 and FO-35 efficiently delivered mRNA to ferret lungs. Overall, this work shows the utility of deep learning for improving nanoparticle delivery.

Lipid nanoparticles (LNPs) have great potential to enable inhaled delivery of mRNA to treat pulmonary diseases. However, this potential has been limited by the challenge of nebulizing the LNPs. Nebulization of LNPs can cause LNPs to aggregate and release encapsulated mRNA, limiting their delivery efficacy. To overcome this challenge, LNPs are developed whereby the PEG-lipid is fully replaced with a zwitterionic polymer (ZIP)-lipid conjugate to greatly enhance the nebulizer stability. LNPs formulated with ZIP-lipids (ZIP-LNPs) were stable to nebulization across a wide range of formulation parameters. The optimized ZIP-LNP formulation, containing reduced cholesterol content relative to traditional PEG-lipid LNPs, demonstrated improved inhaled mRNA delivery in both healthy and mucoobstructed mouse lungs. Repeat administration of the optimized ZIP-LNP formulation was well tolerated and did not result in pulmonary inflammation. This study demonstrates the potential of zwitterionic polymer-lipid conjugates for improving the performance of inhaled mRNA-LNP formulations.

To unlock the full promise of messenger (mRNA) therapies, expanding the toolkit of lipid nanoparticles is paramount. However, a pivotal component of lipid nanoparticle development that remains a bottleneck is identifying new ionizable lipids. Here we describe an accelerated approach to discovering effective ionizable lipids for mRNA delivery that combines machine learning with advanced combinatorial chemistry tools. Starting from a simple four-component reaction platform, we create a chemically diverse library of 584 ionizable lipids. We screen the mRNA transfection potencies of lipid nanoparticles containing those lipids and use the data as a foundational dataset for training various machine learning models. We choose the best-performing model to probe an expansive virtual library of 40,000 lipids, synthesizing and experimentally evaluating the top 16 lipids flagged. We identify lipid 119-23, which outperforms established benchmark lipids in transfecting muscle and immune cells in several tissues. This approach facilitates the creation and evaluation of versatile ionizable lipid libraries, advancing the formulation of lipid nanoparticles for precise mRNA delivery.

Dysfunctions in epigenetic regulation play critical roles in tumor development and progression. Histone deacetylases (HDACs) and histone acetyl transferase (HAT) are functionally opposing epigenetic regulators, which control the expression status of tumor suppressor genes. Upregulation of HDAC activities, which results in silencing of tumor suppressor genes and uncontrolled proliferation, predominates in malignant tumors. Inhibition of the deacetylase activity of HDACs is a clinically validated cancer therapy strategy. However, current HDAC inhibitors (HDACi) have elicited limited therapeutic benefit against solid tumors. Here, we disclosed a class of HDACi that are selective for sub-class I HDACs and preferentially accumulate within the normal liver tissue and orthotopically implanted liver tumors. We observed that these compounds possess exquisite on-target effects evidenced by their induction of dose-dependent histone H4 hyperacetylation without perturbation of tubulin acetylation status and G0/G1 cell cycle arrest. Representative compounds 2 and 3a are relatively non-toxic to mice and robustly suppressed tumor growths in an orthotopic model of HCC as standalone agents. Collectively, our results suggest that these compounds may have therapeutic advantage against HCC relative to the current systemic HDACi. This prospect merits further comprehensive preclinical investigations.

The piperazine heterocycle is broadly exploited in FDA-approved drugs and biologically active compounds, but its chemical diversity is usually limited to ring nitrogen substitutions, leaving the four carbon atoms underutilized. Using an efficient six-step synthesis, chiral amino acids were transformed into 3-substituted piperazine-2-acetic acid esters as diastereomeric mixtures whose cis and trans products (dr 0.56 → 2.2:1, respectively) could be chromatographically separated. From five amino acids (both antipodes) was obtained a complete matrix of 20 monoprotected chiral 2,3-disubstituted piperazines, each as a single absolute stereoisomer, all but one in multigram quantities. In keeping with our overall purpose of constructing more Csp³-enriched compound libraries for drug discovery, these diverse and versatile piperazines can be functionalized on either nitrogen atom, allowing them to be used as scaffolds for parallel library synthesis and as intermediates for the production of novel piperazine compounds.