Department of Computer Science

Network alignment aims to uncover topologically similar regions in the protein-protein interaction (PPI) networks of two or more species under the assumption that topologically similar regions tend to perform similar functions. Although there exist a plethora of both network alignment algorithms and measures of topological similarity, currently no gold standard exists for evaluating how well either is able to uncover functionally similar regions. Here we propose a formal, mathematically and statistically rigorous method for evaluating the statistical significance of shared GO terms in a global, 1-to-1 alignment between two PPI networks. Given an alignment in which k aligned protein pairs share a particular GO term g, we use a combinatorial argument to precisely quantify the p-value of that alignment with respect to g compared to a random alignment. The p-value of the alignment with respect to all GO terms, including their inter-relationships, is approximated using the Empirical Browns Method. We note that, just as with BLASTs p-values, this method is not designed to guide an alignment algorithm towards a solution; instead, just as with BLAST, an alignment is guided by a scoring matrix or function; the p-values herein are computed after the fact, providing independent feedback to the user on the biological quality of the alignment that was generated by optimizing the scoring function. Importantly, we demonstrate that among all GO-based measures of network alignments, ours is the only one that correlates with the precision of GO annotation predictions, paving the way for network alignment-based protein function prediction.

Generative Artificial Intelligence is set to revolutionize healthcare delivery by transforming traditional patient care into a more personalized, efficient, and proactive process. Chatbots, serving as interactive conversational models, will probably drive this patient-centered transformation in healthcare. Through the provision of various services, including diagnosis, personalized lifestyle recommendations, dynamic scheduling of follow-ups, and mental health support, the objective is to substantially augment patient health outcomes, all the while mitigating the workload burden on healthcare providers. The life-critical nature of healthcare applications necessitates establishing a unified and comprehensive set of evaluation metrics for conversational models. Existing evaluation metrics proposed for various generic large language models (LLMs) demonstrate a lack of comprehension regarding medical and health concepts and their significance in promoting patients well-being. Moreover, these metrics neglect pivotal user-centered aspects, including trust-building, ethics, personalization, empathy, user comprehension, and emotional support. The purpose of this paper is to explore state-of-the-art LLM-based evaluation metrics that are specifically applicable to the assessment of interactive conversational models in healthcare. Subsequently, we present a comprehensive set of evaluation metrics designed to thoroughly assess the performance of healthcare chatbots from an end-user perspective. These metrics encompass an evaluation of language processing abilities, impact on real-world clinical tasks, and effectiveness in user-interactive conversations. Finally, we engage in a discussion concerning the challenges associated with defining and implementing these metrics, with particular emphasis on confounding factors such as the target audience, evaluation methods, and prompt techniques involved in the evaluation process.

Most online chemical reaction databases are not publicly accessible or are fully downloadable. These databases tend to contain reactions in noncanonicalized formats and often lack comprehensive information regarding reaction pathways, intermediates, and byproducts. Within the few publicly available databases, reactions are typically stored in the form of unbalanced, overall transformations with minimal interpretability of the underlying chemistry. These limitations present significant obstacles to data-driven applications including the development of machine learning models. As an effort to overcome these challenges, we introduce PMechDB, a publicly accessible platform designed to curate, aggregate, and share polar chemical reaction data in the form of elementary reaction steps. Our initial version of PMechDB consists of over 100,000 such steps. In the PMechDB, all reactions are stored as canonicalized and balanced elementary steps, featuring accurate atom mapping and arrow-pushing mechanisms. As an online interactive database, PMechDB provides multiple interfaces that enable users to search, download, and upload chemical reactions. We anticipate that the public availability of PMechDB and its standardized data representation will prove beneficial for chemoinformatics research and education and the development of data-driven, interpretable models for predicting reactions and pathways. PMechDB platform is accessible online at https://deeprxn.ics.uci.edu/pmechdb.

The subiculum (SUB), a hippocampal formation structure, is among the earliest brain regions impacted in Alzheimer's disease (AD). Toward a better understanding of AD circuit-based mechanisms, we mapped synaptic circuit inputs to dorsal SUB using monosynaptic rabies tracing in the 5xFAD mouse model by quantitatively comparing the circuit connectivity of SUB excitatory neurons in age-matched controls and 5xFAD mice at different ages for both sexes. Input-mapped brain regions include the hippocampal subregions (CA1, CA2, CA3), medial septum and diagonal band, retrosplenial cortex, SUB, postsubiculum (postSUB), visual cortex, auditory cortex, somatosensory cortex, entorhinal cortex, thalamus, perirhinal cortex (Prh), ectorhinal cortex, and temporal association cortex. We find sex- and age-dependent changes in connectivity strengths and patterns of SUB presynaptic inputs from hippocampal subregions and other brain regions in 5xFAD mice compared with control mice. Significant sex differences for SUB inputs are found in 5xFAD mice for CA1, CA2, CA3, postSUB, Prh, lateral entorhinal cortex, and medial entorhinal cortex: all of these areas are critical for learning and memory. Notably, we find significant changes at different ages for visual cortical inputs to SUB. While the visual function is not ordinarily considered defective in AD, these specific connectivity changes reflect that altered visual circuitry contributes to learning and memory deficits. Our work provides new insights into SUB-directed neural circuit mechanisms during AD progression and supports the idea that neural circuit disruptions are a prominent feature of AD.

This project focuses on the development of a human-aware advanced driver assistance system (ADAS) that helps promote safe driving based on a driver's biometrics and driving behavior. The system uses biometric signals from the driver using BioHarness Belt and Galvanic Skin Response (GSR) sensors to monitor the driver's heart rate, breathing rate, ECG and GSR signals. These sensory information are then analyzed by machine learning models to determine whether the driver is stressed or drowsy. The project extended the current state-of-the-art driving simulator CARLA to not only include the driver's brake intensity, speed, throttle and steering but also the driver's biometric state. The vehicle controls and biometrics are then plotted and analyzed for correlations that could imply a driver is driving aggressively, assertively, or defensively. Based on these correlations, the driver will be displayed a warning on the simulation which advises them to drive carefully. The output from this project is an extension to the CARLA tool that developers can use to design and simulate human-aware solutions for ADAS. Faculty Advisor: Professor Salma Elmalaki

The overall objective for HyperXite 9 was to design and build a more robust, and reliable pod, capable of proving the feasibility of a high-speed transportation system. We are working to improve a linear induction motor as the pod's propulsion system. We are also designing and implementing a thermal cooling system to actively dissipate the heat generated by this propulsion system. Our team is comprised of the following 7 subteams: Static Structures, Braking & Pneumatics, Dynamic Structures, Propulsion, Power Systems, Control Systems, and Outreach.

As the field of computational cognitive neuroscience continues to expand and generate new theories, there is a growing need for more advanced methods to test the hypothesis of brain-behavior relationships. Recent progress in Bayesian cognitive modeling has enabled the combination of neural and behavioral models into a single unifying framework. However, these approaches require manual feature extraction, and lack the capability to discover previously unknown neural features in more complex data. Consequently, this would hinder the expressiveness of the models. To address these challenges, we propose a Neurocognitive Variational Autoencoder (NCVA) to conjoin high-dimensional EEG with a cognitive model in both generative and predictive modeling analyses. Importantly, our NCVA enables both the prediction of EEG signals given behavioral data and the estimation of cognitive model parameters from EEG signals. This novel approach can allow for a more comprehensive understanding of the triplet relationship between behavior, brain activity, and cognitive processes.

Animals in the natural world constantly encounter geometrically complex landscapes. Successful navigation requires that they understand geometric features of these landscapes, including boundaries, landmarks, corners and curved areas, all of which collectively define the geometry of the environment1-12. Crucial to the reconstruction of the geometric layout of natural environments are concave and convex features, such as corners and protrusions. However, the neural substrates that could underlie the perception of concavity and convexity in the environment remain elusive. Here we show that the dorsal subiculum contains neurons that encode corners across environmental geometries in an allocentric reference frame. Using longitudinal calcium imaging in freely behaving mice, we find that corner cells tune their activity to reflect the geometric properties of corners, including corner angles, wall height and the degree of wall intersection. A separate population of subicular neurons encode convex corners of both larger environments and discrete objects. Both corner cells are non-overlapping with the population of subicular neurons that encode environmental boundaries. Furthermore, corner cells that encode concave or convex corners generalize their activity such that they respond, respectively, to concave or convex curvatures within an environment. Together, our findings suggest that the subiculum contains the geometric information needed to reconstruct the shape and layout of naturalistic spatial environments.

Stacking and twisting graphene layers allows to create and control a two-dimensional electron liquid with strong correlations. Experiments indicate that these systems exhibit strong tendencies towards both magnetism and triplet superconductivity. Motivated by this phenomenology, we study a 2D model of fluctuating triplet pairing and spin magnetism. Individually, their respective order parameters, d and N, cannot order at finite temperature. Nonetheless, the model exhibits a variety of vestigial phases, including charge-4e superconductivity and broken time-reversal symmetry. Our main focus is on a phase characterized by finite d ⋅ N, which has the same symmetries as the BCS state, a Meissner effect, and metastable supercurrents, yet rather different spectral properties: most notably, the suppression of the electronic density of states at the Fermi level can resemble that of either a fully gapped or nodal superconductor, depending on parameters. This provides a possible explanation for recent tunneling experiments in the superconducting phase of graphene moiré systems.

Electrodeposited amorphous hydrated iridium oxide (IrOx) is a promising material for pH sensing due to its high sensitivity and the ease of fabrication. However, durability and variability continue to restrict the sensors effectiveness. Variation in probe films can be seen in both performance and fabrication, but it has been found that performance variation can be controlled with potentiostatic conditioning (PC). To make proper use of this technique, the morphological and chemical changes affecting the conditioning process must be understood. Here, a thorough study of this material, after undergoing PC in a pH-sensing-relevant potential regime, was conducted by voltammetry, scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDS), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS). Fitting of XPS data was performed, guided by raw trends in survey scans, core orbitals, and valence spectra, both XPS and UPS. The findings indicate that the PC process can repeatably control and conform performance and surface bonding to desired calibrations and distributions, respectively; PC was able to reduce sensitivity and offset ranges to as low as ±0.7 mV/pH and ±0.008 V, respectively, and repeat bonding distributions over ~2 months of sample preparation. Both Ir/O atomic ratios (shifting from 4:1 to over 4.5:1) and fitted components assigned hydroxide or oxide states based on the literature (low-voltage spectra being almost entirely with suggested hydroxide components, and high-voltage spectra almost entirely with suggested oxide components) trend across the polarization range. Self-consistent valence, core orbital, and survey quantitative trends point to a likely mechanism of ligand conversion from hydroxide to oxide, suggesting that the conditioning process enforces specific state mixtures that include both theoretical Ir(III) and Ir(IV) species, and raising the conditioning potential alters the surface species from an assumed mixture of Ir species to more oxidized Ir species.