Artificial neural networks are increasingly used for geophysical modeling to extract complex nonlinear patterns from geospatial data. However, it is difficult to understand how networks make predictions, limiting trust in the model, debugging capacity, and physical insights. EXplainable Artificial Intelligence (XAI) techniques expose how models make predictions, but XAI results may be influenced by correlated features. Geospatial data typically exhibit substantial autocorrelation. With correlated input features, learning methods can produce many networks that achieve very similar performance (e.g., arising from different initializations). Since the networks capture different relationships, their attributions can vary. Correlated features may also cause inaccurate attributions because XAI methods typically evaluate isolated features, whereas networks learn multifeature patterns. Few studies have quantitatively analyzed the influence of correlated features on XAI attributions. We use a benchmark framework of synthetic data with increasingly strong correlation, for which the ground truth attribution is known. For each dataset, we train multiple networks and compare XAI-derived attributions to the ground truth. We show that correlation may dramatically increase the variance of the derived attributions, and investigate the cause of the high variance: is it because different trained networks learn highly different functions or because XAI methods become less faithful in the presence of correlation? Finally, we show XAI applied to superpixels, instead of single grid cells, substantially decreases attribution variance. Our study is the first to quantify the effects of strong correlation on XAI, to investigate the reasons that underlie these effects, and to offer a promising way to address them.

Background: Candida auris is an opportunistic pathogen reported in the US since 2016. C. auris infections (CAI) are frequently healthcare-associated, but only one case of donor-derived CAI in a lung transplant recipient has been reported (PMID 28520901). We describe a cluster of two CAIs at a single center in South Carolina occurring in 2 different recipients from the same solid organ transplant donor. Methods: We describe two cases of invasive CAIs occurring in an academic medical center without prior CAI in Charleston, SC in October 2023. C. auris was identified using Bruker MALDI-TOF and confirmed by the state health department. Results: Patient 1: 40-49 year-old male underwent heart transplantation on day 19 from admission complicated by growth of C. auris on post-op day #15 from a drain. He was readmitted post-op days 22-63 with positive blood cultures for C. auris and underwent re-operation with debridement and hardware removal. C. auris pericarditis required multiple returns to the OR (Figure). He was discharged on micafungin/posaconazole with plans for long-term antifungal therapy. Patient 2: 50-59 year-old male underwent liver and kidney transplantation on day 25 from admission from the same donor as Patient 1 in a separate hospital complex. His course was complicated by possible infected biloma not amendable to drainage and C. auris/C. glabrata fungemia, which was further complicated by abdominal wall collection cultures growing C. auris on post-operative day 35 on washout. He was managed with dual micafungin/posaconazole however, he died of unrelated causes at 93 days after transplant (Figure). Investigation: The donor for both recipients was a 10-19 year-old male who suffered brain death after trauma and was hospitalized for 56 days prior to procurement in Atlanta, GA without known CAI. Airway cultures at the time of organ procurement were positive for rare Pseudomonas and light unidentified yeast of multiple morphologies; urine cultures also grew 40,000 cfu/ml un-identified yeast. Screening of 35 and 4 inpatients in units exposed to patients 1 & 2, respectively, with axilla/groin PCR was negative (Figure). A third organ recipient for this donor (kidney) at our center was identified and had negative urine fungus cultures. Conclusions: Despite no definitive link to a known donor infection, this cluster of CAI occurring simultaneously in 2 patients in separate hospitals/units at a single center with no known prior cases represents likely donor-derived CAI. Our experience suggests that organ procurement organizations should consider improved screening techniques for C. auris in donor cultures.

Disclosure: Scott Curry: Consultant- Ferring; Abbott Diagnostics- Speaker honorarium

Clinical outcomes of repetitive transcranial magnetic stimulation (rTMS) for treatment of treatment-resistant depression (TRD) vary widely and there is no mood rating scale that is standard for assessing rTMS outcome. It remains unclear whether TMS is as efficacious in older adults with late-life depression (LLD) compared to younger adults with major depressive disorder (MDD). This study examined the effect of age on outcomes of rTMS treatment of adults with TRD. Self-report and observer mood ratings were measured weekly in 687 subjects ages 16–100 years undergoing rTMS treatment using the Inventory of Depressive Symptomatology 30-item Self-Report (IDS-SR), Patient Health Questionnaire 9-item (PHQ), Profile of Mood States 30-item, and Hamilton Depression Rating Scale 17-item (HDRS). All rating scales detected significant improvement with treatment; response and remission rates varied by scale but not by age (response/remission ≥ 60: 38%–57%/25%–33%; <60: 32%–49%/18%–25%). Proportional hazards models showed early improvement predicted later improvement across ages, though early improvements in PHQ and HDRS were more predictive of remission in those < 60 years (relative to those ≥ 60) and greater baseline IDS burden was more predictive of non-remission in those ≥ 60 years (relative to those < 60). These results indicate there is no significant effect of age on treatment outcomes in rTMS for TRD, though rating instruments may differ in assessment of symptom burden between younger and older adults during treatment.

Complex machine learning architectures and high-dimensional gridded input data are increasingly used to develop high-performance geoscience models, but model complexity obfuscates their decision-making strategies. Understanding the learned patterns is useful for model improvement or scientific investigation, motivating research in eXplainable artificial intelligence (XAI) methods. XAI methods often struggle to produce meaningful explanations of correlated features. Gridded geospatial data tends to have extensive autocorrelation so it is difficult to obtain meaningful explanations of geoscience models. A recommendation is to group correlated features and explain those groups. This is becoming common when using XAI to explain tabular data. Here, we demonstrate that XAI algorithms are highly sensitive to the choice of how we group raster elements. We demonstrate that reliance on a single partition scheme yields misleading explanations. We propose comparing explanations from multiple grouping schemes to extract more accurate insights from XAI. We argue that each grouping scheme probes the model in a different way so that each asks a different question of the model. By analyzing where the explanations agree and disagree, we can learn information about the scale of the learned features. FogNet, a complex three-dimensional convolutional neural network for coastal fog prediction, is used as a case study for investigating the influence of feature grouping schemes on XAI. Our results demonstrate that careful consideration of how each grouping scheme probes the model is key to extracting insights and avoiding misleading interpretations.

The spatial distribution of in situ sessile organisms, including those from the fossil record, provides information about life histories, such as possible dispersal and/or settlement mechanisms, and how taxa interact with one another and their local environments. At Nilpena Ediacara National Park (NENP), South Australia, the exquisite preservation and excavation of 33 fossiliferous bedding planes from the Ediacara Member of the Rawnsley Quartzite reveals in situ communities of the Ediacara Biota. Here, the spatial distributions of three relatively common taxa, Tribrachidium, Rugoconites, and Obamus, occurring on excavated surfaces were analyzed using spatial point pattern analysis. Tribrachidium have a variable spatial distribution, implying that settlement or post-settlement conditions/preferences had an effect on populations. Rugoconites display aggregation, possibly related to their reproductive methods in combination with settlement location availability at the time of dispersal and/or settlement. Additionally, post-settlement environmental controls could have affected Rugoconites on other surfaces, resulting in lower populations and densities. Both Tribrachidium and Rugoconites also commonly occur as individuals or in low numbers on a number of beds, thus constraining possible reproductive strategies and environmental/substrate preferences. The distribution of Obamus is consistent with selective settlement, aggregating near conspecifics and on substrates of mature microbial mat. This dispersal process is the first example of substrate-selective dispersal among the Ediacara Biota, thus making Obamus similar to numerous modern sessile invertebrates with similar dispersal and settlement strategies.

Structural racism in the USA has roots that extend deep into healthcare and medical research, and it remains a key driver of illness and early death for Black, Indigenous, People of Color (BIPOC). Furthermore, the persistence of racism within academic medicine compels an interrogation of education and research within this context. In the spirit of this interrogation, this article highlights a unique model of community-engaged education that integrates cultural humility. As an individual and institutional stance, cultural humility denotes lifelong learning and self-critique, the mitigation of power imbalances, and accountability. The integration of cultural humility emphasizes that when space is created for BIPOC communities to lead the way, education regarding healthcare and research can be effectively reimagined. Demonstrating this effectiveness, six community partners led the development and implementation of a five-module Structural Racism in Healthcare and Research course. Using a cohort model approach, the pilot course enrolled 12 community members and 12 researchers. The curriculum covered topics such as history of racism in healthcare and research, and introduced participants to a cultural resilience framework. Evaluation results demonstrated a significant increase in participants’ knowledge and ability to identify and take action to address inequities related to racism in healthcare and research.

Trace fossils record foraging behaviors, the search for resources in patchy environments, of animals in the rock record. Quantification of the strength, density, and nature of foraging behaviors enables the investigation of how these may have changed through time. Here, we present a novel approach to explore such patterns using spatial point process analyses to quantify the scale and strength of ichnofossil spatial distributions on horizontal bedding planes. To demonstrate the utility of this approach, we use two samples from the terminal Ediacaran Shibantan Member in South China (between 551 and 543 Ma) and the early Cambrian Nagaur Sandstone in northwestern India (between 539 and 509 Ma). We find that ichnotaxa on both surfaces exhibited significant nonhomogeneous lateral patterns, with distinct levels of heterogeneity exhibited by different types of trace fossils. In the Shibantan, two ichnotaxa show evidence for mutual positive aggregation over a shared resource, suggesting the ability to focus on optimal resource areas. Trace fossils from the Nagaur Sandstone exhibit more sophisticated foraging behavior, with greater niche differentiation. Critically, mark correlation functions highlight significant spatial autocorrelation of trace fossil orientations, demonstrating the greater ability of these Cambrian tracemakers to focus on optimal patches. Despite potential limitations, these analyses hint at changes in the development and optimization of foraging at the Ediacaran/Cambrian transition and highlight the potential of spatial point process analysis to tease apart subtle differences in behavior in the trace fossil record.

Constraining patterns of growth using directly observable and quantifiable characteristics can reveal a wealth of information regarding the biology of the Ediacara biota—the oldest macroscopic, complex community-forming organisms in the fossil record. However, these rely on individuals captured at an instant in time at various growth stages, and so different interpretations can be derived from the same material. Here we leverage newly discovered and well-preserved Dickinsonia costata Sprigg, 1947 from South Australia, combined with hundreds of previously described specimens, to test competing hypotheses for the location of module addition. We find considerable variation in the relationship between the total number of modules and body size that cannot be explained solely by expansion and contraction of individuals. Patterns derived assuming new modules differentiated at the anterior result in numerous examples in which the oldest module(s) must decrease in size with overall growth, potentially falsifying this hypothesis. Observed polarity as well as the consistent posterior location of defects and indentations support module formation at this end in D. costata. Regardless, changes in repeated units with growth share similarities with those regulated by morphogen gradients in metazoans today, suggesting that these genetic pathways were operating in Ediacaran animals.

Vulture populations are in severe decline across Africa and prioritization of geographic areas for their conservation is urgently needed. To do so, we compiled three independent datasets on vulture occurrence from road-surveys, GPS-tracking, and citizen science (eBird), and used maximum entropy to build ensemble species distribution models (SDMs). We then identified spatial vulture conservation priorities in Ethiopia, a stronghold for vultures in Africa, while accounting for uncertainty in our predictions. We were able to build robust distribution models for five vulture species across the entirety of Ethiopia, including three Critically Endangered, one Endangered, and one Near Threatened species. We show that priorities occur in the highlands of Ethiopia, which provide particularly important habitat for Bearded Gypaetus barbatus, Hooded Necrosyrtes monachus, Rüppell’s Gyps rüppelli and White-backed Gyps africanus Vultures, as well as the lowlands of north-eastern Ethiopia, which are particularly valuable for the Egyptian Vulture Neophron percnopterus. One-third of the core distribution of the Egyptian Vulture was protected, followed by the White-backed Vulture at one-sixth, and all other species at one-tenth. Overall, only about one-fifth of vulture priority areas were protected. Given that there is limited protection of priority areas and that vultures range widely, we argue that measures of broad spatial and legislative scope will be necessary to address drivers of vulture declines, including poisoning, energy infrastructure, and climate change, while considering the local social context and aiding sustainable development.

Gravitational waves from coalescing neutron stars encode information about nuclear matter at extreme densities, inaccessible by laboratory experiments. The late inspiral is influenced by the presence of tides, which depend on the neutron star equation of state. Neutron star mergers are expected to often produce rapidly rotating remnant neutron stars that emit gravitational waves. These will provide clues to the extremely hot post-merger environment. This signature of nuclear matter in gravitational waves contains most information in the 2–4 kHz frequency band, which is outside of the most sensitive band of current detectors. We present the design concept and science case for a Neutron Star Extreme Matter Observatory (NEMO): a gravitational-wave interferometer optimised to study nuclear physics with merging neutron stars. The concept uses high-circulating laser power, quantum squeezing, and a detector topology specifically designed to achieve the high-frequency sensitivity necessary to probe nuclear matter using gravitational waves. Above 1 kHz, the proposed strain sensitivity is comparable to full third-generation detectors at a fraction of the cost. Such sensitivity changes expected event rates for detection of post-merger remnants from approximately one per few decades with two A+ detectors to a few per year and potentially allow for the first gravitational-wave observations of supernovae, isolated neutron stars, and other exotica.

Previous genetic association studies have failed to identify loci robustly associated with sepsis, and there have been no published genetic association studies or polygenic risk score analyses of patients with septic shock, despite evidence suggesting genetic factors may be involved. We systematically collected genotype and clinical outcome data in the context of a randomized controlled trial from patients with septic shock to enrich the presence of disease-associated genetic variants. We performed genomewide association studies of susceptibility and mortality in septic shock using 493 patients with septic shock and 2442 population controls, and polygenic risk score analysis to assess genetic overlap between septic shock risk/mortality with clinically relevant traits. One variant, rs9489328, located in AL589740.1 noncoding RNA, was significantly associated with septic shock (p = 1.05 × 10–10); however, it is likely a false-positive. We were unable to replicate variants previously reported to be associated (p < 1.00 × 10–6 in previous scans) with susceptibility to and mortality from sepsis. Polygenic risk scores for hematocrit and granulocyte count were negatively associated with 28-day mortality (p = 3.04 × 10–3; p = 2.29 × 10–3), and scores for C-reactive protein levels were positively associated with susceptibility to septic shock (p = 1.44 × 10–3). Results suggest that common variants of large effect do not influence septic shock susceptibility, mortality and resolution; however, genetic predispositions to clinically relevant traits are significantly associated with increased susceptibility and mortality in septic individuals.

The authors report on 7Li, 19F, and 1H pulsed field gradient NMR measurements of 26 organosilyl nitrile solvent-based electrolytes of either lithium bis(trifluorosulfonyl)imide (LiTFSI) or lithium hexafluorophosphate. Lithium transport numbers (as high as 0.50) were measured and are highest in the LiTFSI electrolytes. The authors also report on solvent blend electrolytes of fluoroorganosilyl (FOS) nitrile solvent mixed with ethylene carbonate (EC) and diethyl carbonate. Solvent diffusion measurements on an electrolyte with 6% FOS suggest both the FOS and EC solvate the lithium cation. By comparing lithium transport and transference numbers, the authors find less ion pairing in FOS nitrile carbonate blend electrolytes and difluoroorganosilyl nitrile electrolytes.

Drawing on a landscape analysis of existing data-sharing initiatives, in-depth interviews with expert stakeholders, and public deliberations with community advisory panels across the U.S., we describe features of the evolving medical information commons (MIC). We identify participant-centricity and trustworthiness as the most important features of an MIC and discuss the implications for those seeking to create a sustainable, useful, and widely available collection of linked resources for research and other purposes.

OBJECTIVES/SPECIFIC AIMS: Therapeutic hypothermia (TH) is a neuroprotective therapy regularly used in newborn infants following traumatic births. The infant’s temperature is maintained at 33.5°C for 72 hours by a cooling blanket upon which the infant is placed. Parents are not permitted to hold their infant while TH is ongoing due to concerns for unintentional rewarming or accidental dislodging of catheters or other monitoring equipment. Our prior qualitative research with nurse and parent interviews described the inability to hold an infant during TH as a significant source of stress. We assessed the feasibility of a 30-minute period of maternal holding for infants being actively treated with TH and assessed both the maternal experience of holding and the nurse experience of supporting holding. METHODS/STUDY POPULATION: This was a feasibility study employing a mixed-methods approach. Inclusion criteria were gestational age at birth of 35 weeks or greater, absence of clinical or electrographic seizures during the first 24 hours of TH, and designation as “clinically stable” by the attending neonatologist with the infant on room air, nasal cannula, or continuous positive airway pressure. Quantitative data were obtained from vital sign monitoring every 2 minutes before, during and after holding and from maternal and nurse research surveys. Qualitative data were obtained from nurse surveys. Infant rewarming was prevented through use of a thin foam insulating barrier placed between mother and infant during holding. Adverse events were defined as a change in infant temperature greater than 0.5°C above or below 33.5°C, accidental dislodging of central lines/disruption of EEG leads or early termination of holding due vital sign instability present for greater than 2 recorded measurements including infant bradycardia defined as heart rate less than 80 beats per minute, hypotension defined as mean arterial pressure less than 40 mmHg or oxygen saturation of less than 93%. RESULTS/ANTICIPATED RESULTS: There were 10 newborn infants undergoing TH for neonatal encephalopathy (median gestational age 39.4 weeks) and their mothers (median age=31 years) were recruited. Infants remained on the hypothermia blanket during holding and were transferred safely to their mother’s arms without medical equipment malfunction/dislodgement. Holding occurred at a median of 47 hours of life. The mean temperature prior to holding was 33.4°C and at completion of holding the mean temperature was 33.5°C (p=0.18). There were no significant bradycardia, hypotension or oxygen desaturation events. In total, 80% of mothers reported difficulty bonding with their baby prior to holding and 90% reported a high level of stress before holding. After holding, all mothers felt their bond was “stronger” or “much stronger” and all felt “less stressed” or “much less stressed.” After holding, 75% of nurses reported that they felt a more positive emotional response to the infant. One nurse stated, “being a part of this emotional experience made me feel closer and more connected to this family and gave me a different perspective on just what they had been dealing with and feeling since giving birth to their child.” In free text responses, on 5 separate occasions, nurses commented on the relaxed, calmed or less irritable appearance of the infant while being held during TH. DISCUSSION/SIGNIFICANCE OF IMPACT: In this sample of term infants treated with TH, a 30-minute period of maternal holding was not associated with increased temperature or other adverse events. Holding during TH was associated with extremely positive feedback from mothers and nurses. Future larger studies could consider assessing the impact of holding on endocrinological markers of stress and bonding, on infant glycemic control, on breastfeeding success rates, and the impact of earlier and improved bonding on the developmental outcomes of children held during their treatment with TH. Increasing the duration of holding and allowing both parents to hold on more than one occasion during the 72 hours of TH may increase the proposed benefits of this intervention.