Search for: All records

Machine learning models for electronic structure prediction are commonly trained on large-scale datasets in order to achieve high accuracy [1]. However, recent studies in other domains suggest that a substantial portion of training data may be redundant, and smaller, information-rich subsets can yield comparable model performance [2, 3]. The extent to which electronic structure data can be pruned without sacrificing accuracy remains unexplored. In this work, we systematically analyze redundancy in electronic structure datasets across multiple material systems and evaluate the effectiveness of several data pruning strategies, including gradient-based data selection (GraNd)[2], coverage-centric coreset selection (CCS)[3], and random pruning. The GraNd ranks training samples according to the norm of the loss gradient, while CCS employs a density-based distribution cover to identify informative subsets. Our results demonstrate that over half of the data can be pruned with negligible loss in predictive accuracy using any of the three methods. However, at higher pruning ratios, GraNd exhibits a rapid degradation in performance, whereas CCS maintains remarkable accuracy, consistently outperforming random pruning. Additionally, energy predictions derived from the ML-predicted electron density-based confirm that pruning up to two orders of magnitude is feasible while preserving chemical accuracy [4]. These findings reveal a high degree of redundancy in electronic structure datasets and highlight coverage-based data selection as a powerful strategy for enabling data-efficient machine learning models in electronic structure prediction. REFERENCES [1] Pathrudkar, S., Taylor, S., Keripale, A., Gangan, A.S., Thiagarajan, P., Agarwal, S., Marian, J., Ghosh, S. and Banerjee, A.S., 2024. Electronic structure prediction of medium and high entropy alloys across composition space. npj Computational Materials, 11, 378, (2025). [2] Paul, M., Ganguli, S. and Dziugaite, G.K., 2021. Deep learning on a data diet: Finding important examples early in training. Advances in neural information processing systems, 34, pp.20596-20607. [3] Zheng, H., Liu, R., Lai, F. and Prakash, A., 2022. Coverage-centric coreset selection for high pruning rates. arXiv preprint arXiv:2210.15809. [4] S Hossain, P Thiagarajan, S Pathrudkar, S Taylor, AS Gangan, A. S. Banerjee, S. Ghosh, Surprisingly High Redundancy in Electronic Structure Data. arXiv preprint arXiv:2507.09001 

In response to the growing federal mandates for data management and sharing, the California Digital Library (CDL) and the Association of Research Libraries (ARL) have launched a pilot project to test new workflows at 10 U.S. universities. The project utilizes machine-actionable DMPs (maDMPs) to streamline processes, improve communication, and ensure compliance with federal requirements. This lightning talk will discuss the pilot by exploring three case studies that showcase the use of generative AI for crafting compliant data management plans, maDMPs for deploying computing and storage resources, and tracking research outputs through integration with internal data management platforms. 

Abstract Plasticity to reduce activity is a common way prey evade predators. However, by reducing activity prey often experience lower individual growth rates because they encounter their own prey less often. To overcome this cost, natural selection should not simply favor individuals generating stronger plasticity to reduce activity rates but also selection to resume activity once the threat of predation subsides. If such plasticity is adaptive, it should vary under environmental conditions that generate stronger selection for greater plasticity, such as predator density. Using a mesocosm experiment and observational study with a damselfly-prey/fish-predator system, we show that fish predation exerts selection for greater plasticity in activity rates of damselflies. Such selection allows damselfly activity levels to initially decrease and then rebound when the threat of predation dissipates, potentially helping to ameliorate a hypothesized growth penalty from activity reductions. We also find that the extent of plasticity in activity to the threat of fish predation increases, albeit slightly (r2 = 0.04%–0.063%), as fish densities increase across natural lakes, consistent with the idea that the magnitude of plasticity is shaped by environmental conditions underlying selection. Collectively, these results demonstrate how selection acts to drive adaptive plasticity in a common predator avoidance strategy. 

The Hawaiian islands are recognized as a hotspot of biodiversity, including both surface and subsurface habitats. Recent studies of Hawaiian lava tube fauna have continued to reveal new species. Here, a new species of cave dwelling talitroid amphipod in the genus Spelaeorchestia is described from lava tubes on the island of Hawai i. It is compared with the only other known cave amphipod from the Hawaiian archipelago and with a closely related cave talitroid from Japan in the genus Minamitalitrus. 

With over 14 million people living above 3,500 m, the study of acclimatization and adaptation to high altitude in human populations is of increasing importance, where exposure to high altitude (HA) imposes a blood oxygenation and acid–base challenge. A sustained and augmented hypoxic ventilatory response protects oxygenation through ventilatory acclimatization, but elicits hypocapnia and respiratory alkalosis. A subsequent renally mediated compensatory metabolic acidosis corrects pH toward baseline values, with a high degree of interindividual variability. Differential renal compensation between acclimatizing lowlanders (LL) and Tibetan highlanders (TH; Sherpa) with ascent was previously unknown. We assessed ventilatory and renal acclimatization between unacclimatized LL and TH during incremental ascent from 1,400 m to 4,300 m in age- and sex-matched groups of 15-LL (8F) and 14-TH (7F) of confirmed Tibetan ancestry. We compared respiratory and renally mediated blood acid–base acclimatization (PCO₂, [HCO₃⁻], pH) in both groups before (1,400 m) and following day 8 to 9 of incremental ascent to 4,300 m. We found that following ascent to 4,300 m, LL had significantly lower PCO₂(P<0.0001) and [HCO₃⁻] (P<0.0001), and higher pH (P= 0.0037) than 1,400 m, suggesting respiratory alkalosis and only partial renal compensation. Conversely, TH had significantly lower PCO₂(P< 0.0001) and [HCO₃⁻] (P< 0.0001), but unchanged pH (P= 0.1), suggesting full renal compensation, with significantly lower PCO₂(P= 0.01), [HCO₃⁻] (P< 0.0001) and pH (P= 0.005) than LL at 4,300 m. This demonstration of differential integrative respiratory–renal responses between acclimatizing LL and TH may indicate selective pressure on TH, and highlights the important role of the kidneys in acclimatization. 

Here, we explicitly define a half-cell reaction approach for pH calculation using the electrode couple comprised of the solid-state chloride ion-selective electrode (Cl-ISE) as the reference electrode and the hydrogen ionselective ion-sensitive field effect transistor (ISFET) of the Honeywell Durafet as the hydrogen ion (H+)-sensitive measuring or working electrode. This new approach splits and isolates the independent responses of the Cl-ISE to the chloride ion (Cl−) (and salinity) and the ISFET to H+ (and pH), and calculates pH directly on the total scale (pHEXT total) in molinity (mol (kg-soln)−1) concentration units. We further apply and compare pHEXT total calculated using the half-cell and the existing complete cell reaction (defined by Martz et al. (2010)) approaches using measurements from two SeapHOx sensors deployed in a test tank. Salinity (on the Practical Salinity Scale) and pH oscillated between 1 and 31 and 6.9 and 8.1, respectively, over a six-day period. In contrast to established Sensor Best Practices, we employ a new calibration method where the calibration of raw pH sensor timeseries are split out as needed according to salinity. When doing this, pHEXT total had root-mean squared errors ranging between ±0.0026 and ±0.0168 pH calculated using both reaction approaches relative to pHtotal of the discrete bottle samples (pHdisc total). Our results further demonstrate the rapid response of the Cl-ISE reference to variable salinity with changes up to ±12 (30 min)−1. Final calculated pHEXT total were ≤±0.012 pH when compared to pHdisc total following salinity dilution or concentration. These results are notably in contrast to those of the few in situ field deployments over similar environmental conditions that demonstrated pHEXT total calculated using the Cl-ISE as the reference electrode had larger uncertainty in nearshore waters. Therefore, additional work beyond the correction of variable temperature and salinity conditions in pH calculation using the Cl-ISE is needed to examine the effects of other external stimuli on in situ electrode response. Furthermore, whereas past work has focused on in situ reference electrode response, greater scrutiny of the ISFET as the H+-sensitive measuring electrode for pH measurement in natural waters is also needed. 

Protein farnesylation is a post-translational modification where a 15-carbon farnesyl isoprenoid is appended to the C-terminal end of a protein by farnesyltransferase (FTase). This process often causes proteins to associate with the membrane and participate in signal transduction pathways. The most common substrates of FTase are proteins that have C-terminal tetrapeptide CaaX box sequences where the cysteine is the site of modification. However, recent work has shown that five amino acid sequences can also be recognized, including the pentapeptides CMIIM and CSLMQ. In this work, peptide libraries were initially used to systematically vary the residues in those two parental sequences using an assay based on Matrix Assisted Laser Desorption Ionization–Mass Spectrometry (MALDI-MS). In addition, 192 pentapeptide sequences from the human proteome were screened using that assay to discover additional extended CaaaX-box motifs. Selected hits from that screening effort were rescreened using an in vivo yeast reporter protein assay. The X-ray crystal structure of CMIIM bound to FTase was also solved, showing that the C-terminal tripeptide of that sequence interacted with the enzyme in a similar manner as the C-terminal tripeptide of CVVM, suggesting that the tripeptide comprises a common structural element for substrate recognition in both tetrapeptide and pentapeptide sequences. Molecular dynamics simulation of CMIIM bound to FTase further shed light on the molecular interactions involved, showing that a putative catalytically competent Zn(II)-thiolate species was able to form. Bioinformatic predictions of tetrapeptide (CaaX-box) reactivity correlated well with the reactivity of pentapeptides obtained from in vivo analysis, reinforcing the importance of the C-terminal tripeptide motif. This analysis provides a structural framework for understanding the reactivity of extended CaaaX-box motifs and a method that may be useful for predicting the reactivity of additional FTase substrates bearing CaaaX-box sequences.