About

Gustavo Seabra received his Ph.D. from the Chemistry Department of Kansas State University in 2005, working on the development and applications of quantum mechanical methods for predicting molecular photoionization spectra. He then completed a postdoctoral fellowship at the Quantum Theory Project at the University of Florida, Gainesville, where he focused on implementing hybrid quantum mechanical/molecular mechanics (QM/MM) methods for simulating chemical reactions in complex environments such as enzymes and solutions, enabling the understanding of molecular mechanisms and drug candidate exploration. After his postdoc, he worked as a Chemistry Professor at the Federal University of Pernambuco in Brazil, applying these techniques to enzymes related to Alzheimer’s disease, Dengue, and Zika viruses, including docking techniques and virtual high-throughput screening to identify new drug candidates. In 2019, he returned to the University of Florida as a research associate professor in the College of Pharmacy, Department of Medicinal Chemistry, and the Center for Natural Products, Drug Discovery and Development (CNPD3). His research focuses on the development and application of computational chemistry methods—including Quantum Chemistry, Molecular Dynamics, hybrid QM/MM simulations, Docking and Scoring, Virtual High Throughput Screening, and machine learning—to drug discovery projects.

Research topics

• Chemistry
• Biology
• Biochemistry
• Computer Science
• Organic chemistry
• Combinatorial chemistry
• Information Retrieval
• Artificial Intelligence
• Materials science
• Bioinformatics
• Cell biology
• Data science
• Nanotechnology
• Computational biology
• Pharmacology
• Biophysics

Selected publications

• Molecular Basis of α-Glycine C–H Activation by a Nonheme Fe(II)/2-Oxoglutarate Dioxygenase

  Biochemistry · 2026-04-28

  articleOpen access

  performs this unusual chemistry during mycosporine-like amino acid biosynthesis, converting mono- and disubstituted precursors into palythines and revealing unexpected substrate tolerance. Kinetic isotope effects, detection of a transient hydroxylated intermediate, and glyoxylate byproduct formation support an α-hydroxylation-initiated mechanism. High-resolution crystal structures, complemented by molecular docking, molecular dynamics simulations, and site-directed mutagenesis, define an active-site architecture that positions the glycyl substrate in a near-transition-state geometry. Hybrid QM/MM calculations reveal a low-barrier hydrogen-atom-transfer step followed by hydroxyl rebound and implicate a conserved Trp125 in an electron-transfer network that lowers the activation barrier. Together, these findings establish a mechanistic framework for protein-directed α-glycine C-H activation by nonheme iron enzymes and provide a blueprint for engineering Fe/2-OG dioxygenases to expand the chemical diversity of mycosporines and related natural products.

  PublisherOA PDFDOI

• A virtual screening and molecular dynamics approach in search of novel antibiotic chemotypes

  PLoS ONE · 2026-03-20

  articleOpen access

  Due to the constantly evolving threat of antibiotic resistance, there is a dire need for novel antibacterial agents. Dihydropteroate synthase (DHPS) is a key bacterial enzyme which has been targeted for nearly a century as a means of selective treatment of microbial infections and exhibits two orthosteric binding sites - the p-aminobenzoic acid (pABA) site and the pterin site. The former is the target of sulfonamides, the earliest class of synthetic antibiotics, and its mutant forms have conferred resistance to this drug class, diminishing its utility in the clinic. Conversely, the pterin site, which is highly conserved across bacterial species, is purported to be less tolerant of mutations, rendering it an attractive target for novel antibiotics. Inspired by this, we conducted a large virtual screen of more than 450,000 compounds from commercial databases, identifying compounds 8802 and 7034 as potential pterin-site inhibitors. Compound 8802 was quite attractive as a hit due to the ease of generating analogues, leading to the synthesis of novel compounds LST-1 and LST-2. Rigid docking and molecular dynamics suggested favorable binding of these compounds to the pterin site of DHPS, and compound 8802 exhibited superior antibacterial activity compared to its analogues and 7034. Fluorescence polarization assays did not indicate competitive inhibition of pterin-derived probe binding, and surface plasmon resonance (SPR) suggested these compounds bind very weakly to DHPS, in a nonspecific manner. The in silico assessment of the physicochemical and pharmacological properties predicted a favorable overall profile, indicating that these are suitable leads for further study to improve their activity and determine their precise mode of action.

  PublisherDOI

• Apo2Mol: 3D Molecule Generation via Dynamic Pocket-Aware Diffusion Models

  Proceedings of the AAAI Conference on Artificial Intelligence · 2026-03-14

  articleOpen access

  Deep generative models are rapidly advancing structure-based drug design, offering substantial promise for generating small molecule ligands that bind to specific protein targets. However, most current approaches assume a rigid protein binding pocket, neglecting the intrinsic flexibility of proteins and the conformational rearrangements induced by ligand binding, limiting their applicability in practical drug discovery. Here, we propose Apo2Mol, a diffusion-based generative framework for 3D molecule design that explicitly accounts for conformational flexibility in protein binding pockets. To support this, we curate a dataset of over 24,000 experimentally resolved apo-holo structure pairs from the Protein Data Bank, enabling the characterization of protein structure changes associated with ligand binding. Apo2Mol employs a full-atom hierarchical graph-based diffusion model that simultaneously generates 3D ligand molecules and their corresponding holo pocket conformations from input apo states. Empirical studies demonstrate that Apo2Mol can achieve state-of-the-art performance in generating high-affinity ligands and accurately capture realistic protein pocket conformational changes.

  PublisherDOI

• Discovery and Mechanism of 16–19F, a Novel Synthetic Lethal Inhibitor of the PRMT5•MTA Complex in MTAP-Deleted Cancer Cells

  ACS Chemical Biology · 2025-05-29 · 1 citations

  article

  Protein arginine methyltransferase 5 (PRMT5), which uniquely binds to 5'-methylthioadenosine (MTA) among the PRMT family, is emerging as an attractive epigenetic target for 5'-methylthioadenosine phosphorylase (MTAP)-deleted cancer treatments. Here, we report the discovery of a novel inhibitor 16-19F, which is a potent binder to the PRMT5•MTA, PRMT5•SAH, and PRMT5•SAM complexes and selectively inhibited MTAP-deleted cancer cell growth. Based on transcriptome analysis, we found that kinetochore metaphase signaling and cell cycle control of the chromosomal replication pathway were downregulated after 16-19F treatment in the MDA-MB-231 TNBC cell line. Additionally, we identified a new PRMT5 substrate, MCM7, an important component of DNA helicase, and figured out the potential methylation site Arg219 by site-directed mutagenesis and computational analysis. Moreover, we showed that 16-19F treatment regulated MCM7 localization, which is involved through liquid-liquid phase separation mechanisms, including the formation of stress granules. Together, we discovered a potential novel drug candidate and revealed an unknown mechanism in which PRMT5 methylation altered MCM7 localization by modulating stress granule formation.

  PublisherDOI

• Virtual High-Throughput Screening of Ligands for Disrupting PRMT5/pICLn Interaction in Prostate Cancer Cells

  ACS Medicinal Chemistry Letters · 2025-04-25 · 2 citations

  articleOpen access

  Prostate cancer (PC) is the most commonly diagnosed cancer in men worldwide. While androgen deprivation therapy (ADT) is initially effective, many patients develop resistance, progressing to castration-resistant prostate cancer (CRPC). Recent studies have identified the interaction between PRMT5 (protein arginine methyltransferase 5) and pICLn as a promising therapeutic target, as it promotes the transcription of double-strand break (DSB) repair genes that contribute to therapy resistance. To target this pathway, a screening campaign identified J021-0199 as a potential hit compound that disrupts the PRMT5/pICLn interaction. Biochemical assays demonstrated that J021-0199 binds to the N-terminal TIM barrel domain of PRMT5. In CRPC cell lines (LNCaP and 22Rv1), J021-0199 selectively inhibited cancer cell growth. qPCR analysis further revealed downregulation of DNA damage response (DDR) genes involved in homologous recombination, nonhomologous end joining, and G2 arrest. These results support J021-0199 as a promising lead compound for overcoming resistance in CRPC.

  PublisherOA PDFDOI

• DecoyDB: A Dataset for Graph Contrastive Learning in Protein-Ligand Binding Affinity Prediction

  ArXiv.org · 2025-07-08

  preprintOpen access

  Predicting the binding affinity of protein-ligand complexes plays a vital role in drug discovery. Unfortunately, progress has been hindered by the lack of large-scale and high-quality binding affinity labels. The widely used PDBbind dataset has fewer than 20K labeled complexes. Self-supervised learning, especially graph contrastive learning (GCL), provides a unique opportunity to break the barrier by pre-training graph neural network models based on vast unlabeled complexes and fine-tuning the models on much fewer labeled complexes. However, the problem faces unique challenges, including a lack of a comprehensive unlabeled dataset with well-defined positive/negative complex pairs and the need to design GCL algorithms that incorporate the unique characteristics of such data. To fill the gap, we propose DecoyDB, a large-scale, structure-aware dataset specifically designed for self-supervised GCL on protein-ligand complexes. DecoyDB consists of high-resolution ground truth complexes (less than 2.5 Angstrom) and diverse decoy structures with computationally generated binding poses that range from realistic to suboptimal (negative pairs). Each decoy is annotated with a Root Mean Squared Deviation (RMSD) from the native pose. We further design a customized GCL framework to pre-train graph neural networks based on DecoyDB and fine-tune the models with labels from PDBbind. Extensive experiments confirm that models pre-trained with DecoyDB achieve superior accuracy, label efficiency, and generalizability.

  PublisherOA PDFDOI

• GatorAffinity: Boosting Protein-Ligand Binding Affinity Prediction with Large-Scale Synthetic Structural Data

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-01 · 2 citations

  preprintOpen access

  Protein-ligand binding affinity prediction is a fundamental task in computational drug discovery. Although substantial efforts have been made to enhance prediction accuracy using data-driven approaches, progress remains limited by persistent data scarcity. The widely used PDBbind dataset, for example, contains fewer than 20,000 experimental structures with annotated binding affinities, while a vast number of affinity measurements remain underutilized due to missing structural data. Here, we investigate this untapped potential by curating more than 450,000 synthetic protein-ligand complexes annotated with Kd and Ki values using the Boltz-1 structure prediction model. Building on this unprecedented scale of synthetic data, further augmented with over 1 million synthetic complexes from the recently released SAIR database annotated with IC50 values, we develop GatorAffinity, a geometric deep learning-based scoring function pre-trained on large-scale synthetic data and fine-tuned using high-quality experimental structures from PDBbind. Extensive evaluation on a leak-proof benchmark demonstrates that GatorAffinity significantly outperforms state-of-the-art affinity prediction methods, offering superior accuracy and generalizability. Our findings show that augmenting available experimental data with synthetic complexes can effectively address the data scarcity challenge while maintaining strong predictive reliability. By releasing the pretrained GatorAffinity model and the large-scale synthetic dataset GatorAffinity-DB, we provide a scalable and reproducible foundation for affinity prediction, virtual screening, and broader structure-based drug design applications (https://github.com/AIDD-LiLab/GatorAffinity).

  PublisherOA PDFDOI

• Grassystatin G, a new cathepsin D inhibitor from marine cyanobacteria: discovery, synthesis, and biological characterization

  RSC Medicinal Chemistry · 2025-01-01 · 2 citations

  articleOpen access

  -Me-d-Phe in other grassystatins. To prove the structure and overcome the lack of material for further biological studies and mechanistic characterization, we developed a 3 + 3 convergent synthesis and have accessed the peptide with an overall yield of 19% using standard peptide coupling. As the statine moiety is a known pharmacophore with an inhibitory effect against aspartic proteases, we screened grassystatin G against a panel of human and virus aspartic proteases. In contrast to grassystatins A-F, preferentially targeting CatE over CatD with 18-66-fold selectivity, grassystatin G displayed 2-fold selectivity for CatD over CatE, suggesting that the key structural differences may be exploited for CatD probe design. Docking and molecular dynamics provided insights into the structural features responsible for the selectivity towards CatD. CatD is well-documented to play a role in cancer proliferation and metastasis, particularly in the context of breast cancer. We tested grassystatin G against MDA-MB-231 triple-negative breast cancer cells and demonstrated its cooperative effects with TRAIL. RNA-seq highlighted the potential pathways and molecular mechanisms governed by grassystatin G alone and in combination with TRAIL.

  PublisherOA PDFDOI

• Biosynthesis, Chemical Synthesis, and Pharmacological Evaluation of Lyngbyapeptin A as a GPCR Antagonist of Motilin, Cannabinoid, and Amylin Receptors

  Journal of Natural Products · 2025-10-17

  articleOpen access

  to the GPCR hits were further investigated using molecular modeling.

  PublisherOA PDFDOI

• Discovery of PRMT5 N-Terminal TIM Barrel Ligands from Machine-Learning-Based Virtual Screening

  ACS Omega · 2025-01-02 · 5 citations

  articleOpen access

  Protein arginine methyltransferase 5 (PRMT5), which symmetrically dimethylates cytosolic and nuclear proteins, has been demonstrated as an important cancer therapeutic target. In recent years, many advanced achievements in PRMT5 inhibitor development have been made. Most PRMT5 inhibitors in the clinical trial focus on targeting the C-terminal catalytic domain, whereas developing small molecules to interrupt the PRMT5/pICLn (methylosome subunit) protein–protein interface is also of great importance for inhibiting PRMT5. Here, we describe a machine-learning-based virtual screening method and use this novel pipeline to screen small-molecule inhibitors of the PRMT5/pICLn interaction. 18 compounds were manually selected for experimental testing. One compound, Z319334062, showed surface plasmon resonance-binding affinity to the target (KD = 21.5 μM) and dose-dependently inhibited symmetric dimethylation levels in patient-derived glioblastoma cell lines.

  PublisherDOI

Frequent coauthors

• Adrián E. Roitberg

  University of Florida

  35 shared

• Ross C. Walker

  26 shared

• Chenglong Li

  15 shared

• Lina Cui

  University of Florida

  12 shared

• Jun Liu

  University of Florida Health Science Center

  10 shared

• David A. Case

  Rutgers, The State University of New Jersey

  9 shared

• Jorge Luiz Neves

  Universidade Federal de Pernambuco

  9 shared

• Marcus Elstner

  Karlsruhe Institute of Technology

  9 shared