Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by the presence of repetitive patterns of behaviors and deficits in social communication and interaction. Attention to social information is a key component of the development of social communication. Previous studies used eye tracking to examine visual scanning patterns associated with social attention in children with ASD and neurotypical children. Eye tracking is used to both identify the atypical patterns of social attention and to predict clinical outcomes in ASD. Although atypical eye gazing patterns are considered as potential biomarkers, researchers commonly consider data loss in eye tracking as error or noise, and rarely investigate it more thoroughly. In this proposal, we hypothesize that loss of data is a potential signature of core social motivation issue when a social video is playing, and, rather than being a nuisance variable, which reflects the broader continuum of social attentional-motivational challenges faced by individuals with ASD. We used eye tracking to confirm previous findings on atypical attention patterns, and further utilize behavior coding to examine the three types of causes of data loss including blinking, non-compliant behaviors, and technical error. We hypothesize that data loss due to blinking is associated with a lack of social motivation and that data loss due to non-compliant behaviors is associated with executive function. Social motivation and executive function were measured by parent reports. Exploring data loss in eye tracking may help reveal comprehensive and fundamental factors of diminished social motivation and neurocognition in ASD.

The maternal bond is an intimate attachment between a primary caregiver (PC) and their infant which provides the infant with security, facilitating physical, social, and emotional development. A sensitive and responsive environment, such as the presence of healthy maternal bonds, guides an infant’s neurodevelopment. Changes in mood and emotional state can alter the care a PC provides and cause difficulties in bonding with their infant, impacting the baby’s psychological and physical development. In infants, the mechanisms by which development may be impacted are unknown. Recently, research has shown early atypical attention to visual pop-out in autism spectrum disorders. Attention to visual pop-out describes our cognitive ability to quickly identify differing objects presented among similar looking ones. In this project, we investigated whether PC mental health affects attention to visual pop-out in infants. Participants included 50 infants who were assessed at 6 and 12 months of age. PCs completed the Beck Depression Inventory (BDI) and Beck Anxiety Inventory (BAI) at both timepoints. Infants watched a 4-minute visual pop-out paradigm, which included social (face) and non-social (shape) trials. We assessed responses to visual pop-outs and explored whether BDI and BAI scores correlated with visual pop-out performance at 6 and 12 months. We also investigated whether BDI and BAI scores influenced the development of attention to visual pop-out between 6 and 12 months. We hypothesized infants of primary caregivers who report more (a) depressive and (b) anxious symptoms will demonstrate weaker identification of the pop-out during social trials compared to their peers but be unaffected during nonsocial trials. This study will help deepen our understanding of the impact of maternal depression and anxiety on infant development and help health providers identify and support families.

Automatic gender recognition (AGR) is a subfield of facial recognition that has recently been scrutinized for bias in the form of misgendering and erasure against various identity groups in our society. Recent studies have found that several commercial AGR classifiers (from Microsoft, IMB, Face++) are biased against women and darker-skinned people as well as gender non-binary people. In this work, we investigate and quantify AGR classifier bias against transgender people by developing and evaluating three different convolutional neural networks (CNN): using images of cisgender individuals, using images of transgender individuals, and using images of both cisgender and transgender individuals. We find that the cisgender trained classifier is 91.7% accurate when evaluated on cisgender people, but only 68.9% accurate when evaluated on transgender people, with the worst performance on trans men with 38.6% precision. We also find that the classifier trained on the combined dataset performs nearly as well as and occasionally outperforms both other classifiers when evaluated on their own datasets, highlighting potential methods for avoiding overfitting. Additionally, we visualize how the classifiers differ by obscuring different parts of the face. Overall, the disparities of accuracy between each classifier demonstrate the degree to which they are impacted by the composition on their dataset and highlight the possibility for commercial AGR classifiers potential to misgender trans people, in particular, transgender men, at a high rate.

Immunological memory prevents reinfection by a pathogen. This protection is accomplished by memory T cells expressing T cell receptors (TCR) specific for previously encountered pathogen-derived peptides (antigens). Conventionally, memory T cells are thought to be inert during novel infections because there is no interaction between their TCRs with their specific antigens. Despite this, we and others have demonstrated that these T cells (here termed “bystanders”) can be activated by inflammatory signals alone and gain cytotoxic effector function in the absence of TCR-antigen interaction. This study aims to determine how inflammation regulates and attenuates bystander responses and how we can leverage these cells therapeutically. Using in vitro cell stimulations, we found that the inhibitory receptor, programmed cell death protein 1 (PD-1), is strongly upregulated by bystanders after exposure to certain inflammatory cytokines. This finding is unique because the current paradigm is that PD-1 expression is caused by TCR stimulation and PD-1 represents a target to manipulate bystander responses. Further, in mouse models of vaccination, we found that bystander-mediated killing can limit vaccine antigen. We believe that interfering with bystander T cell effector functionality could be targeted to improve antigen-specific vaccine responses. Through understanding the mechanisms that dictate bystander function, we may better modulate bystander T cells function during infection, vaccination, and cancer to improve patient outcomes.

Escherichia coli (E. coli) bacteria are a source of food related illness. If irrigation water is contaminated by fecal matter runoff, crops may become infected prior to harvesting, processing, or packaging. Existing test methods require 16-48 hours for sufficient growth and subsequent confirmation of bacterial infection in the irrigation water. Therefore, providing a means for a rapid detection of water borne coliform and E. coli during this growth phase would allow a more preventative response. We have developed a method to determine bacteria presence by a measure of metabolic activity with a spectral analysis system. Byproducts of fermentation from the metabolic activity of live bacteria results in a solution pH drop within a relatively short time. The fluorophore fluorescein is added to the media, allowing optical detection of the solution pH due to its pH sensitive spectral properties within a pH range of 4-7. A blue LED is used to excite fluorescence, emitting peaks at 525 and 550 nm wavelength light depending on the ionization state equilibrium. Unmixing of the spectral profile yields the fluorescent contributions of the ionization states and determination of the pH. A pH drop from metabolic activity serves as a confirmatory test for a growing bacteria culture. Results can be provided within the early hours of growth instead of days, with time of detection depending on the initial concentration of living bacteria. The economical and biosafe characteristics of fluorescein and the testing materials would allow the use of the assay in low resource or rural areas.

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will conduct an all-sky survey and uses difference imaging to identify and classify variable sources. Galaxies with an actively accreting supermassive black hole at the center, or active galactic nuclei (AGN), can vary in brightness significantly. They are powerful tools for improving our understanding of high energy astrophysics. We present an analysis of a subset of high-probability AGN from the High Cadence Transient Survey (HiTS) data release, a precursor for LSST. By comparing difference imaging light curves generated by both HiTS and LSST software, we eliminate bad sources, crossmatch with other datasets, and identify previously unknown AGN. We use these comparisons to assess why some known AGN were not found in the HiTS data. We also make suggestions for how the LSST software can be used to maximize variable AGN science.