Iterative Prompt Refinement for Mining Gene Relationships from ChatGPT.

ChatGPT has demonstrated its potential as a surrogate knowledge graph. Trained on extensive data sources, including open-access publications, peer-reviewed research articles and biomedical websites, ChatGPT extracted information on gene relationships and biological pathways. However, a major challenge is model hallucination, i.e., high false positive rates. To assess and address this challenge, we systematically evaluated ChatGPT's capacity for predicting gene relationships using GPT-3.5-turbo and GPT-4. Benchmarking against the KEGG Pathway Database as the ground truth, we experimented with diverse prompting strategies, targeting gene relationships of activation, inhibition, and phosphorylation. We introduced an innovative iterative prompt refinement technique. By assessing prompt efficacy using metrics like F-1 score, precision, and recall, GPT-4 was re-engaged to suggest improved prompts. A refined prompt, which combines a specialized role with explanatory text, significantly enhances the performance. Going beyond pairwise gene relationships, we also deciphered complex gene interplays, such as gene interaction chains and pathways pertinent to diseases like non-small cell lung cancer. Direct prompts showed limited success, but "least-to-most" prompting exhibited significant potentials for such network constructions. The methods in this study may be used for some other bioinformatics prediction problems.

L2pB1 Cells Contribute to Tumor Growth Inhibition.

Natural IgM (nIgM) antibodies play critical roles in cancer immunosurveillance. However, the role of B-1 B cells, the lymphocytes that produce nIgM, remains to be elucidated. L2pB1 cells, a subpopulation of B-1 B cells, have a unique poly-self-reactive nIgM repertoire and are capable of phagocytosis, potent antigen presentation, and immunomodulation. Using an inducible knock-in and knockout mouse model, we investigated the effect of the loss of L2pB1 cells in a B16F10 melanoma model. Our results show active tumor infiltration of L2pB1 cells in wild type mice, and conversely, depletion of L2pB1 cells results in larger tumor mass and increased angiogenesis. In vitro analysis revealed that L2pB1 cells contribute to the growth inhibition of melanoma cells in both 2D cell culture and 3D tumor spheroids. Similar effects were observed in an MC38 murine colon cancer model. Moreover, our data suggest that one of the ways that L2pB1 cells can induce tumor cell death is via lipoptosis. Lastly, we tested whether L2pB1 cell-derived monoclonal nIgM antibodies can specifically recognize tumor spheroids. Nine of the 28 nIgM-secreting L2pB1 clones demonstrated specific binding to tumor spheroids but did not bind control murine embryonic fibroblasts. Our study provides evidence that L2pB1 cells contribute to cancer immunity through their unique nIgM repertoire, tumor recognition, and lipoptosis. Taken together, because of their ability to recognize common features of tumors that are independent of genetic mutations, L2pB1 cells and their nIgM could be potential candidates for cancer treatment that can overcome tumor heterogeneity-associated drug resistance.

Design Characteristics of Geographic Atrophy Treatment Trials: Systematic Review of Registered Trials in ClinicalTrials.gov.

PURPOSE: Currently there is no effective treatment for geographic atrophy (GA) secondary to age-related macular degeneration, despite many trials that were completed or are ongoing. The purpose of this systematic review is to describe the key design characteristics of GA trials. Well-designed GA trials using appropriate outcome measures are critical for the discovery of an effective treatment. This review may provide valuable insights for improving the design of future trials. METHODS: Two reviewers independently searched for eligible trials (last search on February 28, 2017) and abstracted data of design characteristics including eligibility criteria, sample size, length of follow-up, and primary and secondary outcomes. RESULTS: Among 53 registered GA treatment trials, 10 (19%) were in phase 1, 36 (68%) in phase 1-2 or 2 (phase 2), and 7 (13%) in phase 2-3 or 3 (phase 3); 24 of the 53 trials (45%) are ongoing. The most common interventions were anti-inflammatory agents (40%). GA growth was used as the primary outcome in 58% of phase 2 trials and 71% of phase 3 trials, and as a secondary outcome in 31% of phase 2 trials and 0% of phase 3 trials. Visual acuity (VA) was used as the primary outcome in 17% of phase 2 trials and 14% of phase 3 trials, and as a secondary outcome in 67% of phase 2 trials and 57% of phase 3 trials. The median sample size of trials was 22 in phase 1, 60 in phase 2, and 772 in phase 3, with median follow-up length of 0.7, 1.0, and 2.0 years, respectively. The study eye baseline VA criterion was specified in 33 trials (62%) and varied across trials, with a lower limit of 20/50 to 20/400. The criterion for baseline GA size was specified in 14 trials (26%), with minimum GA size 0.5 to 1 disc area. CONCLUSIONS: There exists large variation in the design characteristics of GA treatment trials. GA growth is the most common primary outcome measure, and VA is the most common secondary outcome measure. Future trials should consider these design characteristics to advance the discovery of an effective treatment for GA.