A synergetic cocatalyst for conversion of carbon dioxide, sunlight, and water into methanol.

The conversion of CO2 into liquid fuels, using only sunlight and water, offers a promising path to carbon neutrality. An outstanding challenge is to achieve high efficiency and product selectivity. Here, we introduce a wireless photocatalytic architecture for conversion of CO2 and water into methanol and oxygen. The catalytic material consists of semiconducting nanowires decorated with core-shell nanoparticles, with a copper-rhodium core and a chromium oxide shell. The Rh/CrOOH interface provides a unidirectional channel for proton reduction, enabling hydrogen spillover at the core-shell interface. The vectorial transfer of protons, electrons, and hydrogen atoms allows for switching the mechanism of CO2 reduction from a proton-coupled electron transfer pathway in aqueous solution to hydrogenation of CO2 with a solar-to-methanol efficiency of 0.22%. The reported findings demonstrate a highly efficient, stable, and scalable wireless system for synthesis of methanol from CO2 that could provide a viable path toward carbon neutrality and environmental sustainability.

Highly stable preferential carbon monoxide oxidation by dinuclear heterogeneous catalysts.

Atomically dispersed catalysts have been shown highly active for preferential oxidation of carbon monoxide in the presence of excess hydrogen (PROX). However, their stability has been less than ideal. We show here that the introduction of a structural component to minimize diffusion of the active metal center can greatly improve the stability without compromising the activity. Using an Ir dinuclear heterogeneous catalyst (DHC) as a study platform, we identify two types of oxygen species, interfacial and bridge, that work in concert to enable both activity and stability. The work sheds important light on the synergistic effect between the active metal center and the supporting substrate and may find broad applications for the use of atomically dispersed catalysts.

Mutation-induced shift of the photosystem II active site reveals insight into conserved water channels.

Photosystem II (PSII) is the water-plastoquinone photo-oxidoreductase central to oxygenic photosynthesis. PSII has been extensively studied for its ability to catalyze light-driven water oxidation at a Mn4CaO5 cluster called the oxygen-evolving complex (OEC). Despite these efforts, the complete reaction mechanism for water oxidation by PSII is still heavily debated. Previous mutagenesis studies have investigated the roles of conserved amino acids, but these studies have lacked a direct structural basis that would allow for a more meaningful interpretation. Here, we report a 2.14-Å resolution cryo-EM structure of a PSII complex containing the substitution Asp170Glu on the D1 subunit. This mutation directly perturbs a bridging carboxylate ligand of the OEC, which alters the spectroscopic properties of the OEC without fully abolishing water oxidation. The structure reveals that the mutation shifts the position of the OEC within the active site without markedly distorting the Mn4CaO5 cluster metal-metal geometry, instead shifting the OEC as a rigid body. This shift disturbs the hydrogen-bonding network of structured waters near the OEC, causing disorder in the conserved water channels. This mutation-induced disorder appears consistent with previous FTIR spectroscopic data. We further show using quantum mechanics/molecular mechanics methods that the mutation-induced structural changes can affect the magnetic properties of the OEC by altering the axes of the Jahn-Teller distortion of the Mn(III) ion coordinated to D1-170. These results offer new perspectives on the conserved water channels, the rigid body property of the OEC, and the role of D1-Asp170 in the enzymatic water oxidation mechanism.

High-resolution cryo-electron microscopy structure of photosystem II from the mesophilic cyanobacterium, Synechocystis sp. PCC 6803.

Photosystem II (PSII) enables global-scale, light-driven water oxidation. Genetic manipulation of PSII from the mesophilic cyanobacterium Synechocystis sp. PCC 6803 has provided insights into the mechanism of water oxidation; however, the lack of a high-resolution structure of oxygen-evolving PSII from this organism has limited the interpretation of biophysical data to models based on structures of thermophilic cyanobacterial PSII. Here, we report the cryo-electron microscopy structure of PSII from Synechocystis sp. PCC 6803 at 1.93-Å resolution. A number of differences are observed relative to thermophilic PSII structures, including the following: the extrinsic subunit PsbQ is maintained, the C terminus of the D1 subunit is flexible, some waters near the active site are partially occupied, and differences in the PsbV subunit block the Large (O1) water channel. These features strongly influence the structural picture of PSII, especially as it pertains to the mechanism of water oxidation.

Mapping N- to C-terminal allosteric coupling through disruption of a putative CD74 activation site in D-dopachrome tautomerase.

The macrophage migration inhibitory factor (MIF) protein family consists of MIF and D-dopachrome tautomerase (also known as MIF-2). These homologs share 34% sequence identity while maintaining nearly indistinguishable tertiary and quaternary structure, which is likely a major contributor to their overlapping functions, including the binding and activation of the cluster of differentiation 74 (CD74) receptor to mediate inflammation. Previously, we investigated a novel allosteric site, Tyr99, that modulated N-terminal catalytic activity in MIF through a "pathway" of dynamically coupled residues. In a comparative study, we revealed an analogous allosteric pathway in MIF-2 despite its unique primary sequence. Disruptions of the MIF and MIF-2 N termini also diminished CD74 activation at the C terminus, though the receptor activation site is not fully defined in MIF-2. In this study, we use site-directed mutagenesis, NMR spectroscopy, molecular simulations, in vitro and in vivo biochemistry to explore the putative CD74 activation region of MIF-2 based on homology to MIF. We also confirm its reciprocal structural coupling to the MIF-2 allosteric site and N-terminal enzymatic site. Thus, we provide further insight into the CD74 activation site of MIF-2 and its allosteric coupling for immunoregulation.

Rapid Quantification of Protein Secondary Structure Composition from a Single Unassigned 1D 13C Nuclear Magnetic Resonance Spectrum.

The function of a protein is predicated upon its three-dimensional fold. Representing its complex structure as a series of repeating secondary structural elements is one of the most useful ways by which we study, characterize, and visualize a protein. Consequently, experimental methods that quantify the secondary structure content allow us to connect a protein's structure to its function. Here, we introduce an automated gradient descent-based method we refer to as secondary-structure distribution by NMR that allows for rapid quantification of the protein secondary structure composition of a protein from a single, 1D 13C NMR spectrum without chemical shift assignments. The analysis of nearly 900 proteins with known structure and chemical shifts demonstrates the capabilities of our approach. We show that these results rival alternative techniques such as FT-IR and circular dichroism that are commonly used to estimate secondary structure compositions. The resulting method requires only the primary sequence of the protein and its referenced 13C NMR spectrum. Each residue is modeled in an ensemble of secondary structures with percentage contributions from random coil, α-helix, and ß-sheet secondary structures obtained by minimizing the difference between a simulated and experimental 1D 13C NMR spectrum. The capabilities of the method are demonstrated as applied to samples at natural abundance or enriched in 13C, acquired by either solution or solid-state NMR, and even on low magnetic field benchtop NMR spectrometers. This approach allows for rapid characterization of protein secondary structure across traditionally challenging to characterize states including liquid-liquid phase-separated, membrane-bound, or aggregated states.

Water Ligands Regulate the Redox Leveling Mechanism of the Oxygen-Evolving Complex of the Photosystem II.

Understanding how water ligands regulate the conformational changes and functionality of the oxygen-evolving complex (OEC) in photosystem II (PSII) throughout the catalytic cycle of oxygen evolution remains a highly intriguing and unresolved challenge. In this study, we investigate the effect of water insertion (WI) on the redox state of the OEC by using the molecular dynamics (MD) and quantum mechanics/molecular mechanics (QM/MM) hybrid methods. We find that water binding significantly reduces the free energy change for proton-coupled electron transfer (PCET) from Mn to YZ•, underscoring the important regulatory role of water binding, which is essential for enabling the OEC redox-leveling mechanism along the catalytic cycle. We propose a water binding mechanism in which WI is thermodynamically favored by the closed-cubane form of the OEC, with water delivery mediated by Ca2+ ligand exchange. Isomerization from the closed- to open-cubane conformation at three post-WI states highlights the importance of the location of the MnIII center in the OEC and the orientation of its Jahn-Teller axis to conformational changes of the OEC, which might be critical for the formation of the O-O bond. These findings reveal a complex interplay between conformational changes in the OEC and the ligand environment during the activation of the OEC by YZ•. Analogous regulatory effects due to water ligand binding are expected to be important for a wide range of catalysts activated by redox state transitions in aqueous environments.

Breaking a Molecular Scaling Relationship Using an Iron-Iron Fused Porphyrin Electrocatalyst for Oxygen Reduction.

The design of efficient electrocatalysts is limited by scaling relationships governing trade-offs between thermodynamic and kinetic performance metrics. This ″iron law″ of electrocatalysis arises from synthetic design strategies, where structural alterations to a catalyst must balance nucleophilic versus electrophilic character. Efforts to circumvent this fundamental impasse have focused on bioinspired applications of extended coordination spheres and charged sites proximal to a catalytic center. Herein, we report evidence for breaking a molecular scaling relationship involving electrocatalysis of the oxygen reduction reaction (ORR) by leveraging ligand design. We achieve this using a binuclear catalyst (a diiron porphyrin), featuring a macrocyclic ligand with extended electronic conjugation. This ligand motif delocalizes electrons across the molecular scaffold, improving the catalyst's nucleophilic and electrophilic character. As a result, our binuclear catalyst exhibits low overpotential and high catalytic turnover frequency, breaking the traditional trade-off between these two metrics.

A Resilient Platform for the Discrete Functionalization of Gold Surfaces Based on N-Heterocyclic Carbene Self-Assembled Monolayers.

We describe the synthesis and characterization of a versatile platform for gold functionalization, based on self-assembled monolayers (SAMs) of distal-pyridine-functionalized N-heterocyclic carbenes (NHC) derived from bis(NHC) Au(I) complexes. The SAMs are characterized using polarization-modulation infrared reflectance-absorption spectroscopy, surface-enhanced Raman spectroscopy, and X-ray photoelectron spectroscopy. The binding mode is examined computationally using density functional theory, including calculations of vibrational spectra and direct comparisons to the experimental spectroscopic signatures of the monolayers. Our joint computational and experimental analyses provide structural information about the SAM binding geometries under ambient conditions. Additionally, we examine the reactivity of the pyridine-functionalized SAMs toward H2SO4 and W(CO)5(THF) and verify the preservation of the introduced functionality at the interface. Our results demonstrate the versatility of N-heterocyclic carbenes as robust platforms for on-surface acid-base and ligand exchange reactions.

Mechanism of Action of KL-50, a Candidate Imidazotetrazine for the Treatment of Drug-Resistant Brain Cancers.

Aberrant DNA repair is a hallmark of cancer, and many tumors display reduced DNA repair capacities that sensitize them to genotoxins. Here, we demonstrate that the differential DNA repair capacities of healthy and transformed tissue may be exploited to obtain highly selective chemotherapies. We show that the novel N3-(2-fluoroethyl)imidazotetrazine "KL-50" is a selective toxin toward tumors that lack the DNA repair protein O6-methylguanine-DNA-methyltransferase (MGMT), which reverses the formation of O6-alkylguanine lesions. We establish that KL-50 generates DNA interstrand cross-links (ICLs) by a multistep process comprising DNA alkylation to generate an O6-(2-fluoroethyl)guanine (O6FEtG) lesion, slow unimolecular displacement of fluoride to form an N1,O6-ethanoguanine (N1,O6EtG) intermediate, and ring-opening by the adjacent cytidine. The slow rate of N1,O6EtG formation allows healthy cells expressing MGMT to reverse the initial O6FEtG lesion before it evolves to N1,O6EtG, thereby suppressing the formation of toxic DNA-MGMT cross-links and reducing the amount of DNA ICLs generated in healthy cells. In contrast, O6-(2-chloroethyl)guanine lesions produced by agents such as lomustine and the N3-(2-chloroethyl)imidazotetrazine mitozolomide rapidly evolve to N1,O6EtG, resulting in the formation of DNA-MGMT cross-links and DNA ICLs in healthy tissue. These studies suggest that careful consideration of the rates of chemical DNA modification and biochemical DNA repair may lead to the identification of other tumor-specific genotoxic agents.

ChemSpaceAL: An Efficient Active Learning Methodology Applied to Protein-Specific Molecular Generation.

The incredible capabilities of generative artificial intelligence models have inevitably led to their application in the domain of drug discovery. Within this domain, the vastness of chemical space motivates the development of more efficient methods for identifying regions with molecules that exhibit desired characteristics. In this work, we present a computationally efficient active learning methodology and demonstrate its applicability to targeted molecular generation. When applied to c-Abl kinase, a protein with FDA-approved small-molecule inhibitors, the model learns to generate molecules similar to the inhibitors without prior knowledge of their existence and even reproduces two of them exactly. We also show that the methodology is effective for a protein without any commercially available small-molecule inhibitors, the HNH domain of the CRISPR-associated protein 9 (Cas9) enzyme. To facilitate implementation and reproducibility, we made all of our software available through the open-source ChemSpaceAL Python package.

Benchmarking various nonadiabatic semiclassical mapping dynamics methods with tensor-train thermo-field dynamics.

Accurate quantum dynamics simulations of nonadiabatic processes are important for studies of electron transfer, energy transfer, and photochemical reactions in complex systems. In this comparative study, we benchmark various approximate nonadiabatic dynamics methods with mapping variables against numerically exact calculations based on the tensor-train (TT) representation of high-dimensional arrays, including TT-KSL for zero-temperature dynamics and TT-thermofield dynamics for finite-temperature dynamics. The approximate nonadiabatic dynamics methods investigated include mixed quantum-classical Ehrenfest mean-field and fewest-switches surface hopping, linearized semiclassical mapping dynamics, symmetrized quasiclassical dynamics, the spin-mapping method, and extended classical mapping models. Different model systems were evaluated, including the spin-boson model for nonadiabatic dynamics in the condensed phase, the linear vibronic coupling model for electronic transition through conical intersections, the photoisomerization model of retinal, and Tully's one-dimensional scattering models. Our calculations show that the optimal choice of approximate dynamical method is system-specific, and the accuracy is sensitively dependent on the zero-point-energy parameter and the initial sampling strategy for the mapping variables.

Beyond the "spine of hydration": Chiral SFG spectroscopy detects DNA first hydration shell and base pair structures.

Experimental methods capable of selectively probing water at the DNA minor groove, major groove, and phosphate backbone are crucial for understanding how hydration influences DNA structure and function. Chiral-selective sum frequency generation spectroscopy (chiral SFG) is unique among vibrational spectroscopies because it can selectively probe water molecules that form chiral hydration structures around biomolecules. However, interpreting chiral SFG spectra is challenging since both water and the biomolecule can produce chiral SFG signals. Here, we combine experiment and computation to establish a theoretical framework for the rigorous interpretation of chiral SFG spectra of DNA. We demonstrate that chiral SFG detects the N-H stretch of DNA base pairs and the O-H stretch of water, exclusively probing water molecules in the DNA first hydration shell. Our analysis reveals that DNA transfers chirality to water molecules only within the first hydration shell, so they can be probed by chiral SFG spectroscopy. Beyond the first hydration shell, the electric field-induced water structure is symmetric and, therefore, precludes chiral SFG response. Furthermore, we find that chiral SFG can differentiate chiral subpopulations of first hydration shell water molecules at the minor groove, major groove, and phosphate backbone. Our findings challenge the scientific perspective dominant for more than 40 years that the minor groove "spine of hydration" is the only chiral water structure surrounding the DNA double helix. By identifying the molecular origins of the DNA chiral SFG spectrum, we lay a robust experimental and theoretical foundation for applying chiral SFG to explore the chemical and biological physics of DNA hydration.

Intrinsic electronic conductivity of individual atomically resolved amyloid crystals reveals micrometer-long hole hopping via tyrosines.

Proteins are commonly known to transfer electrons over distances limited to a few nanometers. However, many biological processes require electron transport over far longer distances. For example, soil and sediment bacteria transport electrons, over hundreds of micrometers to even centimeters, via putative filamentous proteins rich in aromatic residues. However, measurements of true protein conductivity have been hampered by artifacts due to large contact resistances between proteins and electrodes. Using individual amyloid protein crystals with atomic-resolution structures as a model system, we perform contact-free measurements of intrinsic electronic conductivity using a four-electrode approach. We find hole transport through micrometer-long stacked tyrosines at physiologically relevant potentials. Notably, the transport rate through tyrosines (105 s-1) is comparable to cytochromes. Our studies therefore show that amyloid proteins can efficiently transport charges, under ordinary thermal conditions, without any need for redox-active metal cofactors, large driving force, or photosensitizers to generate a high oxidation state for charge injection. By measuring conductivity as a function of molecular length, voltage, and temperature, while eliminating the dominant contribution of contact resistances, we show that a multistep hopping mechanism (composed of multiple tunneling steps), not single-step tunneling, explains the measured conductivity. Combined experimental and computational studies reveal that proton-coupled electron transfer confers conductivity; both the energetics of the proton acceptor, a neighboring glutamine, and its proximity to tyrosine influence the hole transport rate through a proton rocking mechanism. Surprisingly, conductivity increases 200-fold upon cooling due to higher availability of the proton acceptor by increased hydrogen bonding.

Substrate-independent activation pathways of the CRISPR-Cas9 HNH nuclease.

A hallmark of tightly regulated high-fidelity enzymes is that they become activated only after encountering cognate substrates, often by an induced-fit mechanism rather than conformational selection. Upon analysis of molecular dynamics trajectories, we recently discovered that the Cas9 HNH domain exists in three conformations: 1) Y836 (which is two residues away from the catalytic D839 and H840 residues) is hydrogen bonded to the D829 backbone amide, 2) Y836 is hydrogen bonded to the backbone amide of D861 (which is one residue away from the third catalytic residue N863), and 3) Y836 is not hydrogen bonded to either residue. Each of the three conformers differs from the active state of HNH. The conversion between the inactive and active states involves a local unfolding-refolding process that displaces the Cα and side chain of the catalytic N863 residue by ∼5 Å and ∼10 Å, respectively. In this study, we report the two largest principal components of coordinate variance of the HNH domain throughout molecular dynamics trajectories to establish the interconversion pathways of these conformations. We show that conformation 2 is an obligate step between conformations 1 and 3, which are not directly interconvertible without conformation 2. The loss of hydrogen bonding of the Y836 side chain in conformation 3 likely plays an essential role in activation during local unfolding-refolding of an α-helix containing the catalytic N863. Three single Lys-to-Ala mutants appear to eliminate this substrate-independent activation pathway of the wild-type HNH nuclease, thereby enhancing the fidelity of HNH cleavage.

Amine Hole Scavengers Facilitate Both Electron and Hole Transfer in a Nanocrystal/Molecular Hybrid Photocatalyst.

A well-known catalyst, fac-Re(4,4'-R2-bpy)(CO)3Cl (bpy = bipyridine; R = COOH) (ReC0A), has been widely studied for CO2 reduction; however, its photocatalytic performance is limited due to its narrow absorption range. Quantum dots (QDs) are efficient light harvesters that offer several advantages, including size tunability and broad absorption in the solar spectrum. Therefore, photoinduced CO2 reduction over a broad range of the solar spectrum could be enabled by ReC0A catalysts heterogenized on QDs. Here, we investigate interfacial electron transfer from Cd3P2 QDs to ReC0A complexes covalently bound on the QD surface, induced by photoexcitation of the QD. We explore the effect of triethylamine, a sacrificial hole scavenger incorporated to replenish the QD with electrons. Through combined transient absorption spectroscopic and computational studies, we demonstrate that electron transfer from Cd3P2 to ReC0A can be enhanced by a factor of ∼4 upon addition of triethylamine. We hypothesize that the rate enhancement is a result of triethylamine possibly altering the energetics of the Cd3P2-ReC0A system by interacting with the quantum dot surface, deprotonation of the quantum dot, and preferential solvation, resulting in a shift of the conduction band edge to more negative potentials. We also observe the rate enhancement in other QD-electron acceptor systems. Our findings provide mechanistic insights into hole scavenger-quantum dot interactions and how they may influence photoinduced interfacial electron transfer processes.

Tools for analyzing protonation states and for tracing proton transfer pathways with examples from the Rb. sphaeroides photosynthetic reaction centers.

Protons participate in many reactions. In proteins, protons need paths to move in and out of buried active sites. The vectorial movement of protons coupled to electron transfer reactions establishes the transmembrane electrochemical gradient used for many reactions, including ATP synthesis. Protons move through hydrogen bonded chains of waters and hydroxy side chains via the Grotthuss mechanism and by proton binding and release from acidic and basic residues. MCCE analysis shows that proteins exist in a large number of protonation states. Knowledge of the equilibrium ensemble can provide a rational basis for setting protonation states in simulations that fix them, such as molecular dynamics (MD). The proton path into the QB site in the bacterial reaction centers (RCs) of Rb. sphaeroides is analyzed by MD to provide an example of the benefits of using protonation states found by the MCCE program. A tangled web of side chains and waters link the cytoplasm to QB. MCCE analysis of snapshots from multiple trajectories shows that changing the input protonation state of a residue in MD biases the trajectory shifting the proton affinity of that residue. However, the proton affinity of some residues is more sensitive to the input structure. The proton transfer networks derived from different trajectories are quite robust. There are some changes in connectivity that are largely restricted to the specific residues whose protonation state is changed. Trajectories with QB•- are compared with earlier results obtained with QB [Wei et. al Photosynthesis Research volume 152, pages153-165 (2022)] showing only modest changes. While introducing new methods the study highlights the difficulty of establishing the connections between protein conformation.

A quantitative assessment of (bacterio)chlorophyll assignments in the cryo-EM structure of the Chloracidobacterium thermophilum reaction center.

Chlorophylls and bacteriochlorophylls are the primary pigments used by photosynthetic organisms for light harvesting, energy transfer, and electron transfer. Many molecular structures of (bacterio)chlorophyll-containing protein complexes are available, some of which contain mixtures of different (bacterio)chlorophyll types. Differentiating these, which sometimes are structurally similar, is challenging but is required for leveraging structural data to gain functional insight. The reaction center complex from Chloroacidobacterium thermophilum has a hybrid (bacterio)chlorophyll antenna system containing both chlorophyll a and bacteriochlorophyll a molecules. The recent availability of its cryogenic electron microscopy (cryo-EM) structure provides an opportunity for a quantitative analysis of their identities and chemical environments. Here, we describe a theoretical basis for differentiating chlorophyll a and bacteriochlorophyll a in a cryo-EM map, and apply the approach to the experimental cryo-EM maps of the (bacterio)chlorophyll sites of the chloroacidobacterial reaction center. The comparison reveals that at ~ 2.2-Å resolution, chlorophyll a and bacteriochlorophyll a are easily distinguishable, but the orientation of the bacteriochlorophyll a acetyl moiety is not; however, the latter can confidently be assigned by identifying a hydrogen bond donor from the protein environment. This study reveals the opportunities and challenges in assigning (bacterio)chlorophyll types in structural biology, the accuracy of which is vital for downstream investigations.

Small, Electron-Donating Substituents Give CO2 Activation by Permethylpentalene Zirconium Amido Complexes the Upper Hand: A DFT Study of Distortion and Interaction.

An insight into factors controlling CO2 activation is necessary to develop molecular systems that utilize CO2 as a chemical feedstock. Two permethylpentalene zirconium cyclopentadienyl (mono)amido complexes, Pn*ZrCp(NR2), were previously assessed for CO2 activation, and a strong dependence on the amido substituent was observed. The R = Me analogue reacted rapidly and quantitatively at room temperature to form the carbamato complex, while the R = Ph species was inert. Here, we investigate the origin of this reactivity difference using DFT and the distortion-interaction model to characterize steric and electronic contributions to the activation barrier. We find that the barrier for CO2 insertion with R = Me (19.1 kcal/mol) is lower than with R = Ph (36.6 kcal/mol), explaining the inertness of the Ph-substituted analogue. The distortion energy trend follows the steric bulk of the amido substituents, and the bulkier Ph-substituted complex has a consistently higher distortion energy along its potential energy surface than that of the Me-substituted complex. The interaction energy trend follows the electronics, and a more electron-donating Me-substituted complex shows a consistently lower interaction energy. The balance of these effects at the corresponding TS gives a reduced activation barrier. Small, electron-donating substituents therefore facilitate CO2 activation in these complexes.

HAC-Net: A Hybrid Attention-Based Convolutional Neural Network for Highly Accurate Protein-Ligand Binding Affinity Prediction.

Applying deep learning concepts from image detection and graph theory has greatly advanced protein-ligand binding affinity prediction, a challenge with enormous ramifications for both drug discovery and protein engineering. We build upon these advances by designing a novel deep learning architecture consisting of a 3-dimensional convolutional neural network utilizing channel-wise attention and two graph convolutional networks utilizing attention-based aggregation of node features. HAC-Net (Hybrid Attention-Based Convolutional Neural Network) obtains state-of-the-art results on the PDBbind v.2016 core set, the most widely recognized benchmark in the field. We extensively assess the generalizability of our model using multiple train-test splits, each of which maximizes differences between either protein structures, protein sequences, or ligand extended-connectivity fingerprints of complexes in the training and test sets. Furthermore, we perform 10-fold cross-validation with a similarity cutoff between SMILES strings of ligands in the training and test sets and also evaluate the performance of HAC-Net on lower-quality data. We envision that this model can be extended to a broad range of supervised learning problems related to structure-based biomolecular property prediction. All of our software is available as an open-source repository at https://github.com/gregory-kyro/HAC-Net/, and the HACNet Python package is available through PyPI.