Background: Different intratumoral microbiotaexist in different tumors and play a crucial function in carcinogenesis. However, whether they impact clinical outcomes in esophageal squamous cell carcinoma (ESCC) and their mechanism remain unclear.Methods: 16S rDNA amplicon sequencing was performed on surgically resected samples from 98 ESCC patients to analyze intratumoral microbiome abundance and composition. Multiplex fluorescent immunohistochemistry staining was used to profile the phenotypes of immune infiltrates in the tumor microenvironment (TME).Results: Patients with higher intratumoral Shannon index had significantly worse surgical outcomes. When patients were divided into short-term survivors and long-term survivors based on the median survival time, both intratumoral alpha-diversity and beta-diversity were found to be significantly inconsistent, and the relative abundance of Lactobacillus and Leptotrichia emerged as the two microorganisms that probably influenced the survival of ESCC patients. Only Lactobacillus in ESCC was validated to significantly worsen patients’ prognoses and to be positively correlated with the Shannon index. Multivariate analysis revealed that the intratumoral Shannon index, the relative abundance of Lactobacillus, and the pathologic tumor–node–metastasis (pTNM) stage were independently associated with patients’ overall survival. Furthermore, the relative abundance of both Lactobacillus and Shannon index was positively correlated with the proportions of PD-L1⁺ epithelial cells (ECs) and tumor-associated macrophages (TAMs). The Shannon index was negatively correlated with the proportions of natural killer (NK) cells in the TME.Conclusions: A high abundance of intratumoral Lactobacillus and bacterial alpha-diversity was associated with the formation of the immunosuppressive TME and predicted poor long-term survival in ESCC patients.

Motivation: Creating publication-quality visualizations is essential for bioinformatics but remains a bottleneck for researchers with limited coding expertise. While Large Language Models (LLMs) are proficient at generating code, they often fail in practice due to library dependencies, dataset mismatches, or syntax errors. These issues require manual intervention, slowing data interpretation. Results: We present ggplotAgent, a novel multi-modal, self-debugging artificial intelligence agent that automates publication-ready ggplot2 visualizations. It features a dual-layered framework that resolves code execution errors and uses a vision-enabled agent to verify aesthetic correctness. In benchmarks against the DeepSeek-V3 model, ggplotAgent achieved a 100% code executability rate(versus 85%) and a “Publication-Ready” score of 1.9 (versus 0.7). Surprisingly, it showcased the ability to act as an expert collaborator by intelligently enhancing plots beyond the user’s literal prompt, achieving a positive Insight Score of +0.3 over than the baseline (−0.05). These results demonstrate its ability to reliably produce accurate, high-quality visualizations directly from natural language. Availability and implementation: ggplotAgent is freely accessible as a public web application at https://ggplotagent.databio1.com/ and an offline Streamlit app. The source code is available on GitHub at https://github.com/charlin90/ggplotAgent. This software is distributed under the MIT License.