Marine phytoplankton account for 50% of global primary productivity, regulate nutrient cycles, and form the base of the oceanic food web. Recent studies on freshwater species have shown that cross-feeding between phytoplankton and bacteria in their microbiome influences phytoplankton thermal tolerance. The phytoplankton supplies bacteria with photosynthate and receives cobalamin in return. This allows phytoplankton to synthesize an essential amino acid, methionine, with a specific pathway (METH) that functions better at high temperatures than an alternate pathway (METE). Few studies have investigated this mutualism in marine species. In this study, we investigate how cross-feeding affects marine phytoplankton and their response to global warming. We developed and analyzed an ODE model where the mutualists exchange the nutrients they synthesize (cobalamin and carbon) while also competing for nitrogen. The model explores tradeoffs between allocating resources to growth VS. substrate synthesis, temperature effects on algal growth, conditions required for coexistence, and sensitivity of the mutualism to warming. Focusing on the ecological interaction, we found that temperature affects the species’ equilibrium population value and nutrient limitation scenario. When we imposed a tradeoff between growth and substrate synthesis that species were allowed to evolve their position along, we found that the stability of this ecological interaction collapsed. We explore accounting for nutrient concentrations within cells to stabilize this interaction from an eco-evolutionary perspective. This study provides a quantitative tool for understanding algal blooms under climate change. Additionally, it expands fundamental understanding of how species interactions affect adaptation to thermal gradients.

Fluorescent imaging and electron microscopy (EM) are powerful tools to visualize components of cells and investigate biological processes. However, fluorophores and EM stains commonly have challenges when probing into the cell such as cell toxicity. Moreover, while correlative light-electron microscopy (CLEM) has both benefits of EM and florescent imaging to accurately visualize cell behaviors, several fluorophores are not stable enough to use in CLEM imaging. This project develops hybrid molecules for dual EM/fluorescent imaging. We are currently developing new synthetic routes that will allow us to tether iodinated boron clusters with existing fluorescent molecules. The use of iodinated boron clusters would potentially allow us to introduce non-toxic and photostable boron clusters into cells, which we could then use as building blocks for EM staining. We were able to synthesize several iodinated boron clusters that enable alkylation, and several spectroscopic techniques such as nuclear magnetic resonance spectroscopy and mass spectrometry confirmed the formation of each product. When covalently attached to a fluorescent dye, this system should provide an ability to correlate the corresponding fluorescent signal arising from the fluorophore with an EM image. Thus, these newly synthesized boron compounds have the potential to enhance the quality of cell imaging.

Deep learning methods are a promising tool for constraining cosmological models from large sky surveys, which provide large amounts of data, including millions of images. However, due to observational limitations, these surveys have fewer galaxies at greater distances (redshifts), creating a substantial selection bias. Using biased observational data in training machine learning models can lead to systematic errors in predicting galaxy distances from new data, which will limit our ability to constrain cosmological models. Using multi-band photometry data from the Hyper-Suprime Cam Survey as a training set for machine learning methods, we tested different re-sampling methods to examine whether they could mitigate the effects of observational bias on predicted redshift values. Under-sampling and over-sampling methods were used so that the training set had a uniform distribution of the number of galaxies as a function of redshift. We utilized spectroscopic redshifts of galaxies from available data to use as true labels. We evaluated the predicted redshifts based on the resulting bias and mean squared error. Over-sampling the training data with Monte Carlo data augmentation improved both metrics compared to a randomly sampled control set. Under-sampling the training data was ineffective at improving upon both metrics compared to the control. Furthermore, having a more uniform data representation in the training set led to improved predictions at higher redshifts. This work shows that is possible to reduce biases in machine learning predictions by over-sampling the existing training data. The results also provided a basis for further investigations into re-sampling methods.

The literature provides a range of theoretical models and empirical estimates for the rates of job offers, turnover, and exogenous job destruction. In practice, the frequency of job offers, and whether workers accept these offers, varies depending on macroeconomic conditions and microeconomic characteristics. We narrow the scope from the entire distribution of workers to those who earn inflation-adjusted hourly wages of \$15 or less to study such potential heterogeneity. Theoretically, the \$15 threshold is a generous definition of minimum wage; practically, workers earning at or below this wage live near the poverty line even if they work full-time. For this population, we hypothesize that earnings loss associated with job loss to be more detrimental than among other workers. Using public data from the U.S. Bureau of the Census, we discuss here qualitative analyses on employment and job displacement trends by educational attainment and job type. We also lay out implications for designing unemployment assistance programs.

20% of the ATP produced by the body is consumed by the brain. As a result, mitochondrial metabolism is tightly regulated during brain activity, and metabolic dysfunction is often observed in chronic neurological disease. Although it would be useful to study metabolism in neurons from animal models of chronic neurodegeneration, isolation of neurons in older animals is exceedingly difficult. However, synaptosomes are synaptic nerve terminals that include mitochondria, synaptic vesicles, and postsynaptic densities, and it can be isolated from old animals. Therefore, we conducted experiments to optimize the isolation and purity of synaptosomes for Seahorse bioenergetic assays and gas chromatography-mass spectrometry (GC-MS). We initially isolated synaptosomes by homogenizing mouse brains and using percoll gradients. Then, we performed the bicinchoninic acid assay (BCA) to optimize the experimental working dynamic range of synaptosomes sample to load in the instrument of GC-MS. Additionally, to optimize the isolated synaptosomes purity, we varied the number of strokes during homogenization to increase the purity of intact synaptosomes. Homogenizing the one hemisphere of the brain with eight light strokes maximized the purity of the intact synaptosomes. We tested the oxygen consumption rate in response to different substrates in the Seahorse XF Analyzer to determine their purity and function. The ability to isolate synaptosomes from brain cells is important for the study of mitochondrial metabolism in brains, and synaptosomes can be used to study metabolism in preclinical models of neurodegenerative disease.