ImmunoAIzer

ImmunoAIzer analyzes the spatial distribution of tumor-infiltrating lymphocytes (TILs) and cancer cells within the tumor microenvironment (TME) and detects tumor gene mutations to inform cancer immunotherapy decisions and prognostic evaluation.

Key Features:

• Cellular Biomarker Distribution Prediction (CBDPN): A deep learning, semi-supervised Cellular Biomarker Distribution Prediction Network predicts spatial distributions of CD3, CD20, PanCK, and DAPI within the TME with reported accuracy of 90.4%.
• Tumor Gene Mutation Detection (TGMDN): Using CBDPN-identified tumor regions on hematoxylin and eosin (H&E) slides, a Multilabel Tumor Gene Mutation Detection Network detects mutations in APC, KRAS, and TP53 with AUCs of 0.76, 0.77, and 0.79 respectively.

Scientific Applications:

• Colon cancer TME analysis: Integrates spatial cellular biomarker prediction and gene mutation detection to characterize tumor microenvironment features in colon cancer samples.
• Immunotherapy guidance: Provides cellular and genetic TME information to support selection or evaluation of cancer immunotherapy strategies.
• Prognostic assessment and personalization: Combines biomarker spatial patterns and mutation status to inform prognostic evaluation and potential personalized treatment approaches.

Methodology:

ImmunoAIzer employs deep learning with two components: a semi-supervised CBDPN trained to predict spatial distributions of CD3, CD20, PanCK, and DAPI; CBDPN predictions are used to identify tumor regions on H&E slides; a Multilabel TGMDN is then trained to detect APC, KRAS, and TP53 mutations, with reported CBDPN accuracy 90.4% and TGMDN AUCs 0.76, 0.77, and 0.79.