Search for: All records

Abstract The Burkholderia cepacia complex (BCC) comprises marine bacteria that secrete Burkholderia cepacia lipase (BCL), an enzyme that hydrolyzes ester-linked substrates. BCL is enriched at the ocean–air interface and can be transferred to the atmosphere via sea spray aerosols (SSAs), which contain salts, organic matter, and microorganisms. SSAs influence the Earth’s radiative balance through cloud formation and can contribute to respiratory exposure. Because BCL has been linked to severe infections in immunocompromised individuals, its presence in SSAs may have implications for both climate-relevant processes and human health. Here, we quantify water evaporation from saline solutions with and without BCL beneath marine-relevant lipid monolayers. Lipid monolayers suppress evaporation relative to pure salt solutions, whereas the addition of BCL restores water loss to near–salt-solution levels, indicating disruption of the interfacial barrier. To probe the molecular basis of this effect, we performed classical all-atom molecular dynamics simulations of BCL in marine-relevant lipid monolayer and bilayer systems, followed by simulations of full-scale, heterogeneous nascent SSA models. Simplified model slab systems reveal BCL conformational dynamics and interfacial stability, while realistic aerosol models reproduce experimental trends and highlight the roles of curvature and lipid packing in modulating water loss. These results demonstrate that structural complexity critically shapes protein behavior at air–water interfaces and emphasize the need for realistic aerosol models in studies of marine interfacial processes. 

We investigate knowledge retrieval with multi-modal queries, i.e. queries containing information split across image and text inputs, a challenging task that differs from previous work on cross-modal retrieval. We curate a new dataset called ReMuQ for benchmarking progress on this task. ReMuQ requires a system to retrieve knowledge from a large corpus by integrating contents from both text and image queries. We introduce a retriever model “ReViz” that can directly process input text and images to retrieve relevant knowledge in an end-to-end fashion without being dependent on intermediate modules such as object detectors or caption generators. We introduce a new pretraining task that is effective for learning knowledge retrieval with multimodal queries and also improves performance on downstream tasks. We demonstrate superior performance in retrieval on two datasets (ReMuQ and OK-VQA) under zero-shot settings as well as further improvements when finetuned on these datasets. 

Bacterial meningitis, an infection of the membranes (meninges) and cerebrospinal fluid (CSF) surrounding the brain and spinal cord, is one of the major causes of death and disability worldwide. Higher case-fatality rates and short survival times have been reported in developing countries. Hence, a quick, straightforward, and low-cost approach is in great demand for the diagnosis of meningitis. In this research, a microfluidic fully paper-based analytical device (μFPAD) integrated with loop-mediated isothermal amplification (LAMP) and ssDNA-functionalized graphene oxide (GO) nano-biosensors was developed for the first time for a simple, rapid, low-cost, and quantitative detection of the main meningitis-causing bacteria, Neisseria meningitidis ( N. meningitidis ). The results can be successfully read within 1 hour with the limit of detection (LOD) of 6 DNA copies per detection zone. This paper device also offers versatile functions by providing a qualitative diagnostic analysis ( i.e. , a yes or no answer), confirmatory testing, and quantitative analysis. These features make the presented μFPAD capable of a simple, highly sensitive, and specific diagnosis of N. meningitis . Furthermore, this microfluidic approach has great potential in the rapid detection of a wide variety of different other pathogens in low-resource settings. 

Knowledge-based visual question answering (VQA) requires answering questions with external knowledge in addition to the content of images. One dataset that is mostly used in evaluating knowledge-based VQA is OK-VQA, but it lacks a gold standard knowledge corpus for retrieval. Existing work leverage different knowledge bases (e.g., ConceptNet and Wikipedia) to obtain external knowledge. Because of varying knowledge bases, it is hard to fairly compare models’ performance. To address this issue, we collect a natural language knowledge base that can be used for any VQA system. Moreover, we propose a Visual Retriever-Reader pipeline to approach knowledge-based VQA. The visual retriever aims to retrieve relevant knowledge, and the visual reader seeks to predict answers based on given knowledge. We introduce various ways to retrieve knowledge using text and images and two reader styles: classification and extraction. Both the retriever and reader are trained with weak supervision. Our experimental results show that a good retriever can significantly improve the reader’s performance on the OK-VQA challenge. 

Water systems often contain complex macromolecular systems that absorb light. In marine environments, these light absorbing components are often at the air–water interface and can participate in the chemistry of the atmosphere in ways that are poorly understood. Understanding the photochemistry and photophysics of these systems represents a major challenge since their composition and structures are not unique. In this study, we present a successful microscopic model of this light absorbing macromolecular species termed “marine derived chromophoric dissolved organic matter” or “m-CDOM” in water. The approach taken involves molecular dynamics simulations in the ground state using on the fly Density Functional Tight-Binding (DFTB) electronic structure theory; Time Dependent DFTB (TD-DFTB) calculations of excited states, and experimental measurements of the optical absorption spectra in aqueous solution. The theoretical hydrated model shows key features seen in the experimental data for a collected m-CDOM sample. As will be discussed, insights from the model are: (i) the low-energy A-band (at 410 nm) is due to the carbon chains combined with the diol- and the oxy-groups present in the structure; (ii) the weak B-band (at 320–360 nm) appears due to the contribution of the ionized speciated form of m-CDOM; and (iii) the higher-energy C-band (at 280 nm) is due to the two fused ring system. Thus, this is a two-speciated formed model. Although a relatively simple system, these calculations represent an important step in understanding light absorbing compounds found in nature and the search for other microscopic models of related materials remains of major interest. 

GQA (CITATION) is a dataset for real-world visual reasoning and compositional question answering. We found that many answers predicted by the best vision-language models on the GQA dataset do not match the ground-truth answer but still are semantically meaningful and correct in the given context. In fact, this is the case with most existing visual question answering (VQA) datasets where they assume only one ground-truth answer for each question. We propose Alternative Answer Sets (AAS) of ground-truth answers to address this limitation, which is created automatically using off-the-shelf NLP tools. We introduce a semantic metric based on AAS and modify top VQA solvers to support multiple plausible answers for a question. We implement this approach on the GQA dataset and show the performance improvements.