Sze, A. | bioRxiv Science

bioRxiv2024

Evaluation of search-enabled Pre-trained Large Language Models on retrieval tasks for the PubChem Database

Databases are indispensable in biological and biomedical research, hosting vast amounts of structured and unstructured data, facilitating the organization, retrieval, and analysis of complex data. Database access, however, remains a manual, tedious, and sometimes overwhelming, task. We investigate in this study the current state of using pre-trained, search-enabled LLMs for data retrieval from biological databases. Equipped with internet search and code generation capabilities, LLMs promise to streamline database access through natural language, expedite search and knowledge retrieval, and provide coherent analytical summaries. As an example database, we focus on evaluating a current search-enabled LLMs (GPT-4o) for retrieval from the PubChem database, a flagship, heavily used database that plays a critical role in biological and biomedical research. As PubChem is an open archival repository, it provides a well-documented programmatic interface that can be exploited through LLM code generation capabilities. We evaluate retrieval tasks for eight common PubChem access protocols that were previously documented. The tasks include identifying interacting genes and proteins, finding drug-like compounds based on structural similarity, retrieving bioactivity data, and locating stereoisomers and isotopomers. We develop a methodology for adopting the protocols into an LLM-prompt, where we supplement the prompt with additional context through iterative prompt refinement as needed. To further evaluate the LLM capabilities, we instruct the LLM to perform the retrieval with and without using programmatic access. We compare the results (referred to as gold and silver answers) when using these retrieval modalities with two traditional retrieval baselines that include running the manual search steps for each reference protocol through the PubChem database web interface, and through the provided PUG (Power-User Gateway) programmatic access. We quantitatively and qualitatively summarize our results, showing that generating programmatic access is more likely to yield the correct answers. We highlight the value and limitations of using current search-based LLMs for database retrieval. We also provide guidance for the future development that can improve the accuracy and reliability of search-based LLMs.

bioinformatics↗

bioRxiv2024

Multi-modal, Label-free, Optical Mapping of Cellular Metabolic Function and Oxidative Stress in 3D Engineered Brain Tissue Models

Brain metabolism is essential for the function of organisms. While established imaging methods provide valuable insights into brain metabolic function, they lack the resolution to capture important metabolic interactions and heterogeneity at the cellular level. Label-free, two-photon excited fluorescence imaging addresses this issue by enabling dynamic metabolic assessments at the single-cell level without manipulations. In this study, we demonstrate the impact of spectral imaging on the development of rigorous intensity and lifetime label-free imaging protocols to assess dynamically metabolic functions over time in 3D engineered brain tissue models comprised of human induced neural stem cells, astrocytes, and microglia. Specifically, we rely on multi-wavelength spectral imaging to identify the excitation/emission profiles of key cellular fluorophores within human brain cells, including NAD(P)H, LipDH, FAD, and lipofuscin. These enable the development of methods to mitigate lipofuscins overlap with NAD(P)H and flavin autofluorescence to extract reliable optical metabolic function metrics from images acquired at two excitation wavelengths over two emission bands. We present fluorescence intensity and lifetime metrics reporting on redox state, mitochondrial fragmentation, and NAD(P)H binding status in neuronal monoculture and the triculture systems to highlight the functional impact of metabolic interactions between different cell types. Our findings reveal significant metabolic differences between neurons and glial cells, shedding light on metabolic pathway utilization, including the glutathione pathway, OXPHOS, glycolysis, and fatty acid oxidation. Collectively, our studies establish a label-free, non-destructive approach to assess the metabolic function and interactions among different brain cell types relying on endogenous fluorescence and illustrate the complementary nature of the information that is gained by combining intensity and lifetime-based images. Such methods can improve understanding of physiological brain function and dysfunction that occurs at the onset of cancers, traumatic injuries and neurodegenerative diseases.

bioengineering↗