Computational and Structural Biotechnology Journal

Pediatric high-grade glioma (pHGG) is a highly aggressive brain tumor characterized by transcriptional plasticity and an immunosuppressive microenvironment. Single-cell RNA-seq analysis revealed diverse malignant and immune cell populations, with tumor-associated macrophages (TAMs) emerging as the primary source of SPP1 (osteopontin), a glycoprotein that suppresses T cell activation through CD44 binding. Cell-cell communication analysis identified the SPP1-CD44 axis as a dominant immunosuppressive pathway in the tumor microenvironment. Despite extensive transcription factor screening, no strong regulators of SPP1 were identified, suggesting regulation occurs via alternative mechanisms. To assess structural features of SPP1, replica exchange molecular dynamics simulations were performed, revealing that the CD44-binding domain is conformationally stable. Phosphorylation at Ser169, a conserved site, further stabilized this region, suggesting a potential mechanism for enhanced CD44 interaction. To disrupt this axis, among 2,500 variants of anti-SPP1 23C3 antibody, a lead candidate variant (23C3-v1) with improved SPP1 binding affinity and minimal sequence divergence was identified. Furthermore, humanized 23C3-v1 (Hu23C3-v1) was designed to neutralize Class I and Class II epitope hits from murine antibody derived 23C3-v1 antibody. Together, this study integrates transcriptomic and structural bioinformatics approaches to the target SPP1-CD44 axis, which can help reduce immunosuppressive characteristics of pHGG tumor microenvironment (TME).

Short linear motif (SLiM)-mediated protein-protein interactions play important roles in several biological processes where transient binding is needed. They usually reside in intrinsically disordered regions (IDRs), which makes them accessible for interaction. Although information about the possible necessity of the flanking regions surrounding the motifs is increasingly available, it is still unclear if there are any generic amino acid attributes that need to be functionally preserved in these segments. Here, we describe the currently known ligand-binding SLiMs and their flanking regions with biologically relevant residue features and analyse them based on their simplified characteristics. Our bioinformatics analysis reveals several important properties in the widely diverse motif environment that presumably need to be preserved for proper motif function, but remained hidden so far. Our results will facilitate the understanding of the evolution of SLiMs, while also hold potential for expanding and increasing the precision of current motif prediction methods. Author summaryProtein-protein interactions between short linear motifs and their binding domains play key roles in several molecular processes. Mutations in these binding sites have been linked to severe diseases, therefore, the interest in the motif research field has been dramatically increasing. Based on the accumulated knowledge, it became evident that not only the short motif sequences themselves, but their surrounding flanking regions also play crucial roles in motif structure and function. Since most of the motifs tend to be located within highly variable disordered protein regions, searching for functionally important physico-chemical properties in their proximity could facilitate novel discoveries in this field. Here we show that the investigation of the motif flanking regions based on different amino acid attributes can provide further information on motif function. Based on our bioinformatics approach we have found so far hidden features that are generally present within certain motif categories, thus could be used as additional information in motif searching methods as well.

Designing next-generation Chimeric Antigen Receptors (CARs) requires a systematic understanding of intracellular signaling domains and their downstream biological effects, yet no comprehensive knowledge resource currently exists for this purpose. Here, we present an automated workflow that integrates multiple natural language processing and large language model tools to extract biomolecular interactions from PubMed literature and assemble them into a CAR T cell signaling knowledge graph. Our pipeline combines REACH, INDRA, and Llama 3 across 15 targeted search queries, yielding a directed multi-relational graph of [~]7,500 unique interactions among [~]1,800 entities, including proteins, biological processes, and chemicals. We further demonstrate that queries incorporating biological process ontology terms retrieve more interaction-rich papers than protein-name-only searches, offering practical guidance for future literature mining efforts. The resulting knowledge base provides a structured foundation for predicting T cell phenotypes and prioritizing intracellular domain candidates for CAR design, with broader applicability to knowledge-driven inference in immunotherapy research.

Alzheimers disease (AD) is characterized by amyloid-{beta} (A{beta}) aggregation, with soluble oligomers implicated as the most neurotoxic species. Recent evidence suggests microbial infections, including Helicobacter pylori, contribute to AD pathogenesis. This study investigates the role of H. pylori GroEL, a conserved chaperonin found in bacterial outer membrane vesicles (OMVs), in stabilizing toxic A{beta} oligomers. A pan-genome analysis of 353 H. pylori strains identified GroEL as a highly conserved protein present in 83% of strains, which supported its widespread relevance. We structurally modelled a conserved 27-amino acid GroEL fragment and docked it against the A{beta} (1-42) tetramer. Interaction analysis revealed stabilizing salt bridges, hydrogen bonds, and extensive non-bonded contacts within the GroEL-A{beta} complex. Molecular dynamics simulations (50 ns) demonstrated that GroEL binding enhanced A{beta} oligomer stability, evidenced by reduced structural deviations and a more extensive hydrogen bonding network compared to A{beta} oligomer alone. These computational findings support a novel mechanism whereby H. pylori GroEL directly stabilizes soluble A{beta} oligomers. We hypothesize that this stabilization inhibits their aggregation into plaques while paradoxically prolonging the lifetime of neurotoxic species, potentially increasing neurodegeneration through pathways distinct from canonical amyloid deposition. This highlights the complex role of bacterial proteins in AD and underscores the need for experimental validation of GroEL-A{beta} interactions as a potential therapeutic target.

Cervical intraepithelial neoplasia grade 2 (CIN2) lesions show variable outcomes, and accurate prediction of regression remains a major clinical challenge. We developed an interpretable machine learning pipeline that integrates quantitative histological, clinical, and human papillomavirus (HPV) -genotyping data to predict lesion regression within one and two years. Using panoptic segmentation of routine hematoxylin and eosin (H&E) -stained biopsies, we extracted human-interpretable morphological and immune cell infiltration related features that capture the key histopathological characteristics of CIN2 and identified features that predicted lesion regression. Further, integrating these features to predictive clinical features achieved higher predictive accuracy than clinical variables alone. These findings demonstrate that quantitative, interpretable analysis of H&E histology of routine diagnostic biopsies contains relevant information that predicts the natural history of CIN2 lesions. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=121 SRC="FIGDIR/small/26344510v1_ufig1.gif" ALT="Figure 1"> View larger version (38K): org.highwire.dtl.DTLVardef@e8ac93org.highwire.dtl.DTLVardef@199f7c6org.highwire.dtl.DTLVardef@159ee1dorg.highwire.dtl.DTLVardef@11fc720_HPS_FORMAT_FIGEXP M_FIG Created in BioRender. Lehtonen, O. (2026) https://BioRender.com/rlnkbkp C_FIG

The pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to pose a threat, with an estimated number of deaths exceeding 5 million. SARS-CoV-2 entry into the cell is mediated by its transmembrane spike glycoprotein (S protein), and the angiotensin-converting enzyme 2 (ACE2) receptor on the human cell surface. The extracellular heparan sulphate (EcHS) enhances the S protein binding through a mechanism that is still unknown. Surprisingly, low molecular weight heparin (LMWH) and HS in the disaccharide form (dHS) hinder the S protein binding to ACE2, despite the similarity with EcHS. We investigated the molecular mechanism behind this inhibition through molecular dynamics (MD) simulations to understand the interaction pattern of the heparinoids with S protein and ACE2 receptor.

Metagenomics has revealed disease-associated shifts in microbial taxa and functions in inflammatory bowel disease (IBD) patients. However, the role of genomic variation in gut commensals remains poorly understood. Here, we integrated metagenomic profiling, variant calling, and structural bioinformatics to identify disease-associated variants in the gut microbes. Crohns disease (CD) and ulcerative colitis (UC) showed significant negative associations with Bacteroides uniformis, Bacteroides vulgatus, and Eubacterium rectale. These bacteria exhibited 190,712 single-nucleotide polymorphisms, including 479 CD-specific and 235 UC-specific variants. Variant prioritization identified a CD-specific Val170Leu substitution in the conserved starch-binding domain of the Starch Utilization System D (SusD) protein in B. uniformis. Structural modeling and cyclodextrin docking indicated reduced binding affinity in the mutant, while 200-ns molecular dynamics simulations showed stable ligand retention only in the wild type. These findings suggest that impaired starch metabolism driven by SusD variation may contribute to B. uniformis depletion in CD and demonstrate the value of integrating metagenomics with structural analyses to identify functionally relevant microbial variants.

The spike (S) protein of SARS-CoV-2 effectuates membrane fusion and virus entry into target cells. Its transmembrane domain (TMD) represents a homotrimer of -helices anchoring the spike in the viral envelope. Although S-protein models available to date include the TMD, its precise configuration was given brief consideration. Understanding viral fusion entails realistic TMD models, while no reliable approaches towards predicting the 3D structure of transmembrane (TM) trimers exist. Here, we propose a comprehensive computational framework to model the spike TMD (S-TMD) based solely on its primary structure. First, we performed amino acid sequence pattern matching and compared molecular hydrophobicity potential (MHP) distribution on the helix surface against TM homotrimers with known 3D structures and thus selected the TMD of the tumour necrosis factor receptor 1 (TNFR-1) for subsequent template-based modelling. We then iteratively built an all-atom homotrimer model of S-TMD based on "dynamic MHP portraits" and residue variability motifs. In this model each helix possessed two overlapping interfaces interacting with either of the remaining helices, which include conservative residues I1216, F1220, I1227, M1229, and M1233. Finally, the stability of this and several alternative models (including a recent NMR structure) and a set of mutant forms was tested in all-atom molecular dynamics (MD) simulations in a POPC bilayer mimicking the viral envelope membrane. Unlike other configurations, our model trimer remained extraordinarily tightly packed over a microsecond-range MD and retained its stability when palmitoylated in accordance with experimental data. Palmitoylation had no significant impact on the TMD conformation nor the way in which the lipid bilayer was perturbed in the presence of the trimer. Overall, the resulting model of S-TMD conforms to known basic principles of TM helix packing and will be further used to explore the complex machinery of membrane fusion from a broader perspective beyond the TMD.

Background and ObjectivesNotch homolog 3 (NOTCH3) gene variants were fully penetrant to produce the disease phenotype of CADASIL. Aberrant NOTCH3 protein leads to degeneration of vascular SMCs and pericytes, targeting microcirculation dysfunction and blood-brain barrier (BBB) leakage. MethodsWe evaluated neuroimaging data of forty patients with NOTCH3 gene variants including eighteen missense/insertion mutations in epidermal growth factor repeat (EGF), negative regulatory region (NRR), and disordered region (Dis). We performed an AI-driven pipeline integrating AlphaFold3, Foldseek, and molecular dynamics simulations to elucidate clinical and molecular consequences. ResultsDistinct domain mutations exhibited characteristic patterns: EGFs 1, 2, 13-15, 32 and Dis correlated with microbleeds/macro-bleeds, lacunes, perivascular spaces, and acute cerebral microinfarcts; EGFs 2, 3, 13-15, 25 with disrupted disulfide bonds or binding motif of protein O-glucosyltransferase 1 (POGLUT1) were predicted to undergo greater structural and functional deteriorations in Notch signaling pathways. NRR/Fab (antigen-binding fragment) destabilized dominant motions and single apo-Dis exhibited low structural disorder. Agreement between computational and experimental data for wild-type EGFs/POGLUT1 and C49F, R75Q, R141C mutants suggests testable hypotheses that advance understanding of cerebral small-vessel disease. DiscussionTargeting POGLUT1 to modulate EGF-like domains and using the Fab region to stabilize NRR complexes may be a promising therapeutic approach deserving rigorous exploration.

Emerging studies underscore the promising capabilities of large language model-based chatbots in conducting fundamental bioinformatics data analyses. The recent feature of accepting image-inputs by ChatGPT motivated us to explore its efficacy in deciphering bioinformatics illustrations. Our evaluation with examples in cancer research, including sequencing data analysis, multimodal network-based drug repositioning, and tumor clonal evolution, revealed that ChatGPT can proficiently explain different plot types and apply biological knowledge to enrich interpretations. However, it struggled to provide accurate interpretations when quantitative analysis of visual elements was involved. Furthermore, while the chatbot can draft figure legends and summarize findings from the figures, stringent proofreading is imperative to ensure the accuracy and reliability of the content.

The aggregation of the myocilin olfactomedin (OLF) domain, generally driven by genetic mutations, is the leading cause of primary open-angle glaucoma (POAG). Developing therapeutic strategies requires a detailed understanding its initial unfolding events that expose aggregation-prone regions (APRs). However, it has been a challenge, as the slow conformational dynamics of OLF hinders classical molecular dynamics (MD) simulations from capturing aggregation-prone OLF intermediates. To overcome this, we employed a multi-pronged computational strategy, integrating over 15 {micro}s of simulation time across diverse conditions, including high-temperature, enhanced sampling, chemical denaturation, and simulations of the pathogenic I499F mutant. Our results reveal that OLF unfolding is not random but initiates at specific structural regions pertinent to the terminal blade A and E. Specifically, the blade interfaces between A-B and A-E showed unique regions rich in aromatic/hydrophobic residues as aggregation hotspots. Overall, our simulations proved effective to generate a detailed map of seven distinct APRs. The accuracy of these APRs is partially validated by the close localization of these predicted regions with both previously identified amyloid peptides and the sites of known disease-causing mutations. By scrutinizing the OLF structure and dynamics under different MD settings, our study provides potential molecular targets for developing new therapeutic interventions against POAG.

Helicobacter pylori has been classified as a Group 1 carcinogen by the International Agency for Research on Cancer of the World Health Organization and is one of the most well-established risk factors for gastric cancer. Long-term colonization by H. pylori depends on adhesin-mediated attachment to the gastric mucosa, among which the blood group antigen-binding adhesin BabA is a key surface factor involved in host recognition, tissue tropism, and persistent infection. In this study, we established a structure-guided computational design pipeline to develop compact protein binders targeting functionally relevant epitopes of BabA. First, using experimentally resolved BabA-antibody and BabA-nanobody complex structures as templates, we extracted structural contact residues on BabA through heavy-atom contact analysis, thereby defining antibody-recognition epitopes supported by complex-structure evidence. In addition, sequence-based, structure-based, and evolutionary conservation analyses were integrated to identify candidate functional epitope residues with high antigenicity, strong conservation, and surface-exposed features. On this basis, constrained de novo backbone generation was performed around the prioritized epitope regions, followed by amino acid sequence design and structural back-validation of the candidate binders. Candidate BabA-binder complexes were further evaluated using molecular docking, molecular dynamics simulations, and residue-level interface perturbation analysis to assess interface stability, epitope occupancy, and potential binding hotspots. This workflow enables systematic screening of BabA-targeting binders that may compete with antibody-recognized functional surfaces. Although these candidates still require experimental validation, this study provides a transferable computational framework for designing compact protein binders against pathogen adhesins by integrating experimentally resolved complex-structure resources with computational epitope prioritization based on sequence, conformation, and evolutionary conservation, and establishes a preliminary library of BabA candidate binders for subsequent validation and optimization.

Single-chain variable fragments (scFvs) are widely used in diagnostic and therapeutic applications. These antibody fragments comprise two antibody variable domains connected by a flexible peptide linker whose properties critically influence folding, stability, oligomeric state, and antigen-binding. Therefore, careful linker selection represents a key step in scFv design. Guanylyl Cyclase C (GUCY2C) is a tumor-associated cell surface receptor expressed in gastrointestinal malignancies, including more than 90% of colorectal cancer (CRC) cases across all disease stages. Its restricted physiological expression pattern makes GUCY2C an attractive target for immunotherapy and precision oncology therapies. Here, we investigated the structural and functional consequences of incorporating alternative linker designs into an anti-GUCY2C scFv. Using molecular modeling, protein-protein docking, and molecular dynamics (MD) simulations, we evaluated the conformational stability, interdomain organization, and antigen-binding interactions of each construct. Our results provide a dynamic, structure-based assessment of how linker composition influences GUCY2C recognition and scFv structural behavior. Furthermore, this work establishes a computational framework for the rational optimization of GUCY2C-targeted antibody fragments.

The 3D organisation of the genome provides an intricate relationship between the chromatin architecture and its effects on the functional state of the cell. Recent advances in high-throughput sequencing and chromosome conformation capture technologies elucidated a comprehensive view of chromatin interactions on a genome-wide scale but provides only a 2D representation of how the chromatin is organised inside the cell nucleus. To quantitatively understand the structural alterations and dynamics of chromatin in 3D, we have developed a computational model that not only captures the hierarchical structural organisation but also provides mechanistic insights into the dynamics of spatial rearrangements of chromatin in developing lymphoid lineage cells. From the combination of approaches of polymer physics representing chromatin as a homopolymeric chain and incorporation of the biological information of chromosomal interactions inferred from the Hi-C data, we generated a coarse grained bead-on-a-string polymer model of chromatin to comprehend the mechanisms underlying the differential chromatin architecture. Our study showed that our simulated chromatin structure recapitulates the intrinsic features of chromatin organisation, including the fractal globule nature, compartmentalization, presence of topologically associating domains (TADs), phase separation and spatial preferences of genomic regions in the chromosomal territories. Comparative analyses of these simulated chromatin structures of differentiating B cell stages revealed compartmental switching and changes in the spatial positioning of lineage specific genomic regions. Analysis of the compactness of the switched regions showed insights into their acquired open-closed states for gene regulation and hence governing the cell fate through consequent structural rearrangement. Based on the remarkable performance of our model, we emphasise on its predictive potential by identifying switching of novel regions that demonstrated undergoing structural rearrangement which was subsequently validated through their differential expression patterns in vitro. These results reveal that although the chromatin organisation seems similar in most cell types, it undergoes distinct structural changes for the regulatory role of chromatin in sustaining cell specificity.

The use of an artificial intelligence (AI)-based prediction model of the structure-odor relationship (SOR) has shown great potential in the replacement of human panelists in gas chromatography-olfactometry (GCO). However, the Al-based GCO encounters issues such as poor accuracy, generalization, and practicality, owning to the insufficient feature extraction of odorant molecular structure. The purpose of this study is to the prediction of odor perception categories based on the odorant structure feature extraction by diverse deep neural networks, including molecular graphic convolution neural network (MG-CNN), molecular graph transformer neural network, and atom interaction neural networks. The results of the performance comparison of different feature extractors demonstrate that the MG-CNN model produces the highest accuracy and thus may be most suitable for the SOR prediction. It is hoped that the proposed method can be applied in practice as an auxiliary tool of GCO for the sensory evaluation of key compounds in food ingredients. HighlightsO_LIDifferent deep neural networks were used to predict categorized odor descriptors. C_LIO_LIEnd-to-end-based representation learning was performed for molecular feature extraction. C_LIO_LIMolecular graphs with pre-training of convolution neural networks was most accurate. C_LI Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=82 SRC="FIGDIR/small/488977v2_ufig1.gif" ALT="Figure 1"> View larger version (33K): org.highwire.dtl.DTLVardef@10495a2org.highwire.dtl.DTLVardef@1fbf27eorg.highwire.dtl.DTLVardef@1ed3664org.highwire.dtl.DTLVardef@8e1383_HPS_FORMAT_FIGEXP M_FIG C_FIG

Recent studies have indicated that the human amyloidogenic protein medin is associated with a range of vascular diseases, including aortic aneurysms, vascular dementia, and Alzheimers disease. Medin accumulates in the vasculature with age, leading to endothelial dysfunction through oxidative and nitrative stress and inducing pro-inflammatory activation. Medin is a cleavage product from the C2 domain of MfgE8. The exact mechanism of medin production from MfgE8 is unknown, with crystal structures of homologous C2 domains suggesting that the cleavage sites are buried, requiring a conformational transition for medin production. Molecular dynamics simulations can explore a wide range of conformations, from small-scale bond rotations to large-scale changes like protein folding or ligand binding. This study employed a combination of full-atom and coarse-grained molecular dynamics simulations, along with CONCOORD- and AlphaFold2-generated models, to investigate MfgE8 conformations and their implications for medin cleavage site accessibility. The simulations revealed that MfgE8 tends to adopt a compact conformation with the RGD motif, important for cell attachment within the N-terminal domain, and the medin region in the C-terminal domain close in proximity. Formation of this compact structure is facilitated by interdomain electrostatic interactions that promote stability and in turn decrease the solvent-accessible surface area of the medin region and particularly the C-terminal medin cleavage site. This data enhances current knowledge on medin generation to propose that alterations in local environmental conditions, possibly through changes in glycosylation or other post-translational modifications are required to induce MfgE8 to unfold partially or fully: this would result in enhanced accessibility of the cleavage sites and therefore enable medin generation.

The SARS-CoV-2 Omicron variants present a different behavior compared to the previous variants, all particularly in respect to the Delta variant, as it seems to promote a lower morbidity although being much more contagious. In this perspective, we performed new molecular dynamics (MD) simulations of the various spike RBD/hACE2 complexes corresponding to the WT, Delta and Omicron variants (BA.1 up to BA.4/5) over 1.5 {micro}s timescale. Then, carrying out a comprehensive analysis of residue interactions within and between the two partners, allowed us to draw the profile of each variant by using complementary methods (PairInt, hydrophobic potential, contact PCA). Main results of PairInt calculations highlighted the most involved residues in electrostatic interactions that represent a strong contribution in the binding with highly stable contacts between spike RBD and hACE2 (importance of mutated residues at positions 417, 493 and 498). In addition to the swappable arginine residues (493/498), the apolar contacts made a substantial and complementary contribution in Omicron with the detection of two hydrophobic patches, one of which was correlated with energetic contribution calculations. This study brings new highlights on the global dynamics of spike RBD/hACE2 complexes resulting from the analysis of contact networks and cross-correlation matrices able to detect subtle changes at point mutations. The results of our study are also consistent with alternative approaches such as binding free energy calculations but are more informative and sensitive to transient or low-energy interactions. Nevertheless, the energetic contributions of residues at positions 501 and 505 were in good agreement with hydrophobic interactions measurements. The contact PCA networks could identify the intramolecular incidence of the S375F mutation occurring in all Omicron variants and likely conferring them an advantage in binding stability. Collectively, these data revealed the major differences observed between WT/Delta and Omicron variants at the RBD/hACE2 interface, which may explain the greater persistence of Omicron. Author SummaryThe evolution of SARS-CoV-2 was extremely rapid, leading to the global predominance of Omicron variants, despite the many mutations identified in the spike protein. Some of these were introduced to evade the immune system, but many others were located in the Receptor Binding Domain (RBD) without affecting its efficient binding to hACE2 and preserving the high infectivity of this variant. To unravel the mechanism by which this protein-protein connection remains strong or stable, it is necessary to study the different types of interactions at the atomic level and over time using molecular dynamics (MD) simulations. Indeed, in contrast to crystal or cryo-EM structures providing only a fixed image of the binding process, MD simulations have allowed to unambiguously identify the sustainability of some interactions mediated by key residues of spike RBD. This study could also highlight the interchangeable role of certain residues in compensating for a mutation, which in turn allows the virus to maintain durable binding to the host cell receptor. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=83 SRC="FIGDIR/small/628120v1_ufig1.gif" ALT="Figure 1"> View larger version (35K): org.highwire.dtl.DTLVardef@b2a6c4org.highwire.dtl.DTLVardef@e29044org.highwire.dtl.DTLVardef@6d9835org.highwire.dtl.DTLVardef@123c6f9_HPS_FORMAT_FIGEXP M_FIG Graphical abstract C_FIG

The interphase chromatin structure is extremely complex, precise and dynamic. Experimental methods can only show the frequency of interaction of the various parts of the chromatin. Therefore, it is extremely important to develop theoretical methods to predict the chromatin structure. In this publication, we describe the necessary factors for the effective modeling of the chromatin structure in Drosophila melanogaster. We also compared Monte Carlo with Molecular Dynamic methods. We showed that incorporating black, non-reactive chromatin is necessary for successfully prediction of chromatin structure, while the loop extrusion model or using Hi-C data as input are not essential for the basic structure reconstruction. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=68 SRC="FIGDIR/small/472204v1_ufig1.gif" ALT="Figure 1"> View larger version (10K): org.highwire.dtl.DTLVardef@65d59aorg.highwire.dtl.DTLVardef@1aa9cbforg.highwire.dtl.DTLVardef@18f5cborg.highwire.dtl.DTLVardef@b1034d_HPS_FORMAT_FIGEXP M_FIG C_FIG

In vitro fertilization (IVF) laboratories equipped with timelapse incubators generate vast quantities of sequential embryo images, yet the absence of standardized, annotated databases impedes the development of reproducible computational tools for embryo assessment. Here we describe the construction of a standardized time-lapse imaging database comprising 631 normally fertilized zygotes from 218 treatment cycles, integrating timelapse image sequences, patient clinical records, and embryo developmental outcomes. We further present a gradient boosting decision tree (GBDT) ensemble framework that integrates zygote morphokinetic parameters-continuous time-series features extracted via a validated CNN-based segmentation algorithm (US Patent US11210494B2)-with conventional embryo assessment grades (categorical features per the Istanbul consensus). The fusion framework employs equal-weight initialization followed by iterative residual-decreasing training to optimally combine heterogeneous feature types. Ablation analysis demonstrated that the integrated model achieved an AUC of 0.78, significantly outperforming morphokinetics-only (AUC 0.71) and conventional-only (AUC 0.65) models, confirming the complementary value of the two data modalities. The database and fusion framework provide a reproducible foundation for embryo development assessment and are generalizable to other multimodal data integration tasks in reproductive medicine.

The HIV-1 protease is one of several common key targets of combination drug therapies for human immunodeficiency virus infection and acquired immunodeficiency syndrome (HIV/AIDS). During the progression of the disease, some individual patients acquire -drug resistance due to mutational hotspots on the viral proteins targeted by combination drug therapies. It has recently been discovered that drug-resistant mutations accumulate on the flap region of the HIV-1 protease, which is a critical dynamic region involved in non-specific polypeptide binding during invasion and infection of the host cell. In this study, we utilize machine learning assisted comparative molecular dynamics, conducted at single amino acid site resolution, to investigate the dynamic changes that occur during functional dimerization and polypeptide binding of the main protease. We use a multi-agent machine learning model to identify conserved dynamics of the HIV-1 main protease that are preserved across simian and feline protease orthologs (SIV and FIV). We also investigate changes in dynamics due to common drug-resistant mutations in many patients. We find that a key functional site in the flap region, a solvent-exposed isoleucine (ILE50) and surrounding sites that control flap dynamics is often targeted by drug-resistance mutations, likely leading to malfunctional molecular dynamics affecting the overall flexibility of the flap region. We conclude that better long term patient outcomes may be achieved by designing drugs that target protease regions which are less dependent upon single sites with large functional binding effects.