National Computational Infrastructure

Advances in QSAR are led by two core paradigms, 1) descriptor engineering, where complex fixedlength vectors of compounds are generated and conventional ML methods are applied to those representations and 2) raw chemical inputs (e.g., SMILES, 2D-graph) being provided to deep learning neural network models, which construct their own internal representations of molecules and learn iteratively over them.Here we present the Tsetlin Machine (TM) -which combines the accuracy and easy-use of existing rule-based QSAR ML methods (e.g., RF and XGBoost), the iterative learning aspect of NN algorithms and its intrinsic interpretability.The TM uses teams of finite-state automata which capture frequent patterns as propositional logic (clauses) via reinforcement learning.The benchmarking pipeline presented here demonstrates that TM-QSAR coupled with ECFP4 descriptors frequently performs better than existing rule-based QSAR methods for ROC-AUC, PRC-AUC and PPV, with a high capacity for inter-scaffold generalisation.However, due to the binary nature of TM-QSAR, performance it is currently limited when descretised continuous descriptors are used.TM-QSAR demonstrated particularly impressive classification scores for MOR (ROC-AUC = 0.87, PRC-AUC = 0.77) and CYPA4 (ROC-AUC = 0.92, PRC-AUC = 0.63), when compared to RF and XGBoost.Using TM in combination with substructural fingerprinting descriptors allows for an interpretability suite which can be extracted directly from clauses.Here we detail molecule property maps (TM-MPM) to view atom-wise TM-QSAR bioactivity contributions for single molecules and closed-form WAC scores (Weights Activations Clauses) for descriptor-wise contributions to regions of predicted chemical space.These methods show strong alignment of TM-QSAR interpretations to known ligandprotein interactions of the MOR target and gives non-linear, conditional interpretations for greater predicted bioactivity.Given this combination of accuracy, computational efficiency and interpretability, we provide a basis for TM-QSAR to be explored as a standard methodology in virtual screening toolkits.

ABSTRACT Social licence to operate (SLO) is increasingly prevalent in social performance and environmental management of impactful sectors. To date, however, research into SLO for major infrastructure development remains limited. This study introduces the SLO for Infrastructure Model using Social Exchange Theory (SET) to build on leading SLO models from other major sectors. The model is constructed from a multi‐year Australian study involving over 1500 survey responses, analysed using Partial Least Squares‐Structural Equation Modelling. The study contributes novel insights into the key factors driving SLO in major infrastructure and responds to critical questions and gaps in the broader SLO literature to offer new perspectives on the role of trust. The study highlights the importance of perceived benefits and perceived risks to SLO, not only for their large effect sizes on community acceptance, but also as key mechanisms through which other predictors of SLO boost and transmit their influence onto the community acceptance necessary to earn and maintain a social licence.

621 Background: Effective communication between clinical researchers and participants is vital for successful clinical trials. However, informed consent documents often contain complex language, making it difficult for many patients to fully understand key information and make informed decisions. This comprehension gap impacts patient experience and impedes trial enrollment. Large Language Models (LLMs) show promise for translating specialized medical content into accessible language; however, their effectiveness in clinical trial communication remains underexplored. Multimodal approaches combining LLM-generated content with visual aids may significantly enhance participant understanding and trust in the consent process. This study aims to design and test the feasibility of creating a lay clinical trial summary using LLM technology. Methods: This feasibility pilot leveraged an LLM to extract and customize key clinical trial information from research protocols for prospective participants. An initial literature review leveraged FDA guidance and prior work by Hill et al (2024) to establish a template and delineate content requirements. We utilized Amazon Bedrock with Anthropic's Claude 3.7 to engineer and test prompts. We extracted each information topic separately into phrases or short sentences in JavaScript Object Notation (JSON) and used placeholders to insert content into the template. Clinical experts, communication specialists, and patient advocates reviewed content for accuracy and clarity. Results: Feedback received from patient advocates were incorporated to optimize content relevance, literacy level, and acceptability before producing 2 final outputs for IRB review. Iterative prompt optimization using one-to-two shot examples in conversational, second-person language achieved optimal LLM outputs at 4-6 grade reading level. The template was converted from PowerPoint shapes to a Word table for improved text wrapping, visual presentation, and Section 508 accessibility compliance. Human oversight remained essential for content validation and managing text constraints within fixed cell dimensions. Patient reviewers provided highly positive feedback, endorsing bullet points, simplified study titles, and enhanced typography with wider spacing. Graphics and layout enhancements significantly improved engagement compared to traditional consent materials. Final outputs, after human review, were submitted to IRB for approval. Conclusions: Utilizing an LLM increased efficiency in creating summaries, provided consistency in language and structure, and customized trial content. Implementation required balancing risks of oversimplification against benefits, maintaining human oversight for critical evaluation, and ensuring regulatory compliance. A future study will evaluate participant preference for layperson abstract plus consent form versus consent form alone.

11532 Background: Kaposi sarcoma herpesvirus (KSHV, also known as human herpesvirus 8) is an oncogenic virus that causes Kaposi sarcoma (KS), one of the most prevalent cancers in people with HIV (PWH) worldwide. KSHV also causes primary effusion lymphoma (PEL), a non-Hodgkin lymphoma, a plasmablastic variant of multicentric Castleman disease (MCD), and KSHV-associated inflammatory cytokine syndrome (KICS). KSHV-associated diseases (KADs) can occur alone or concurrently. The major KSHV genetic subtypes A, A5, B, C, D, E and F, based on the K1 open reading frame (ORF), are distributed by global region of origin. A and C genotypes are seen in Europe and North America whereas B and A5 variants are observed in Africa; D, E and F genotypes were described in isolated populations. There are limited data on the association between KSHV subtypes and survival outcomes. Methods: We investigated the ORF K1 subtype, disease characteristics, and impact on survival outcomes in 107 patients (pts) with KADs treated at the HIV/AIDS Malignancy Branch in the United States from 2010-2024. KSHV-DNA was extracted from peripheral blood mononuclear cells (PBMCs), KSHV-positive tissues and body fluids, including effusions. Extracted DNA was tested for KSHV using a CLIA-certified qPCR assay. Samples identified as KSHV-positive by qPCR were sequenced using either next-generation sequencing (NGS) or Sanger sequencing. Results: The cohort consisted of predominantly men (94%) and 55% were Black. Ninety-seven percent of pts had HIV (median CD4 count of 216 cells/mm3 and median HIV viral load 189,590 copies/mL). The most common KAD was KS (92%) followed by PEL (36%), MCD (23%), and KICS (22%). However, 64% had more than one concurrent KAD – 20% had KS and PEL. At the time of analysis, 39 (36%) pts were deceased, 26 (24%) had a PEL diagnosis. Seventy-nine percent of pts were from North America, 11% from Latin America and 9% were from sub-Saharan Africa; KSHV genotypes were consistent with a patient’s geographic origin. Overall, the most common genotype was A (44%), 30% of pts with genotype A had KS alone and 30% had PEL+/-KS (Table). KSHV subtypes based on K1 did not impact survival outcomes (Global Log-rank P=0.9) overall. In 98 pts with KS with and without other KAD, KSHV subtypes did not affect survival but the presence of concurrent PEL as compared to MCD or KICS lead to worse survival in pts [Hazard Ratio: 7.9 (95% confidence interval: 3.4-18.2, P<0.0001)]. Conclusions: In this large cohort of pts with KS and other KAD, KSHV genotype was not associated with survival outcomes. Among pts with KS, concurrent KAD, such as PEL, led to poorer survival. This suggests that clinical manifestations rather than underlying viral variants impacted survival. KSHV genotype prevalence (%) by KAD. Overall % A44% A58% B11% C26% Dual infection 8% KS 30 50 17 29 0 MCD +/- KS 11 0 17 14 22 PEL +/- KS 30 25 42 21 22 KICS + KS 17 25 17 29 33 MCD+PEL+/- KS 13 0 8 7 22

ASAP1 is a multidomain Arf GTPase-activating protein (ArfGAP) that catalyzes GTP hydrolysis on the small GTPase Arf1 and is implicated in cancer progression. The PH domain of ASAP1 enhances its activity greater than 7 orders of magnitude but the underlying mechanisms remain poorly understood. Here, we combined Nuclear Magnetic Resonance (NMR), Molecular Dynamic (MD) simulations and mathematical modeling of functional data to build a comprehensive structural-mechanistic model of the complex of Arf1 and the ASAP1 PH domain on a membrane surface. Our results support a new conceptual model in which the PH domain contributes to efficient catalysis not only by membrane recruitment but by acting as a critical component of the catalytic interface, binding Arf·GTP and allosterically driving it towards the catalytic transition state. We discuss the biological implications of these results and how they may apply more broadly to poorly understood membrane-dependent regulatory mechanisms controlling catalysis of the ArfGAP superfamily as well as other peripheral membrane enzymes.

Abstract Question addressed by the studyA radiological index capable of quantifying pulmonary tuberculosis (TB) disease progression on standard Chest X-Ray (CXR) could enhance clinical management and inform therapeutic decision-making. Our objectives were: to develop a radiological index that quantitatively assesses the progression of pulmonary TB in patients over the course of treatment and follow-up, to evaluate its usefulness by using CXRs from a cohort of TB patients, and to validate it with CXRs from an independent cohort of TB patients.Materials/patients and methodsAn index was developed to identify and classify patterns of radiological findings of both active and inactive pulmonary TB. This index was evaluated in a first cohort comprising a pseudonymized sample of serial CXRs from 21 TB patients at the Perinatal HIV Research Unit (PHRU) in South Africa and validated in a second, larger, independent cohort comprising 50 treated TB patients from the National Institute of Allergy and Infectious Diseases (NIAID) TB Data Portals. The relationship between index values for activity markers and patients’ clinical characteristics was also analysed.ResultsA radiological index, the RUTI-TB Evolution Score (RTBES), was created and validated in two cohorts, achieving 95.96% interobserver agreement (±2 points). In both cohorts, RTBES for active TB signs decreased over follow-up. Higher baseline RTBES correlated with male sex, higher bacterial loads, and extensive pulmonary involvement (including cavitation and pleural effusions). Baseline RTBES≥2 predicted unfavourable outcomes; 80% of failures/deaths occurred above this threshold.Answer to the questionRTBES effectively quantified pulmonary TB severity and tracked treatment evolution on standard CXRs, showing promise as a prognostic tool to inform clinical decision-making.

This study aims to investigate the structural, spectroscopic, and electronic properties of the synthetic missile fuel exo- and endo-tetrahydrodicyclopentadiene (THDCPD, JP-10) using density functional theory (DFT). It is to understand the dominance of the liquid exo-isomer (96%) of the jet fuel from the subtle differences between the isomers. The present DFT calculations reveal that the exo-isomer is 15.51 kJ/mol more stable than the endo-isomer, attributed to the flipping of the triangular ΔC8-C10-C9 ring in its norbornane skeleton. Calculated nuclear magnetic resonance (13C-NMR) and infrared (IR) spectra, validated by experimental data, reveal larger chemical shifts for junction carbons (C1/C2 and C3/C4) due to reduced electron shielding and show distinct vibrational patterns. Charge analysis indicates that all carbon atoms are negatively charged except for the C1/C2 carbons which are positively charged in both isomers. While overall IR spectra of the isomers appear similar, bands near 3000 cm−1 correspond to distinctly different vibrational modes. The exo-isomer’s electronic structure features a more delocalized HOMO and a larger HOMO-LUMO gap (7.63 eV) than the endo-isomer (7.37 eV). All such differences contribute to the properties of exo-THDCPD and, therefore, why the exo-isomer dominates JP-10 fuel.