BostonGene Announces Publication in Cancer Cell Revealing the Development of a Machine Learning Cellular Deconvolution Algorithm that Accurately Reconstructs the Tumor Microenvironment and Supports Oncology Clinical Decision-making
BostonGene today announced the online publication of the manuscript, “Precise reconstruction of the tumor microenvironment using bulk RNA-seq and a unique machine learning algorithm trained on artificial transcriptomes”
WALTHAM, Mass – August 9, 2022
BostonGene today announced the online publication of the manuscript, “Precise reconstruction of the tumor microenvironment using bulk RNA-seq and a unique machine learning algorithm trained on artificial transcriptomes” in Cancer Cell, a premier peer-reviewed scientific journal that publishes high impact results in cancer research and oncology. The study demonstrated the ability of the BostonGene-developed unique and robust machine learning (ML) algorithm named Kassandra to digitally reconstruct the tissue tumor microenvironment (TME) and blood cellular composition, identifying over 50 unique cell populations in total from RNA-seq derived from both archival and fresh tissues. These results indicate that cellular deconvolution can be utilized in future clinical applications to predict blood and tissue microenvironment composition, a critical factor in cancer pathogenesis, clinical outcome, and therapeutic resistance.
Kassandra was trained on a broad collection of > 9,400 tissue and blood-sorted cell RNA profiles incorporated into millions of artificial transcriptomes for this research study. To increase the stability and robustness of BostonGene’s decision-tree ML deconvolution algorithm, bioinformatic correction for technical and biological variability, aberrant cancer cell expression inclusion, and accurate quantification and normalization of transcript expression were applied. Results of the study showed that Kassandra recognizes 51 unique subpopulations and accurately deconvolves less well-characterized stromal and immune elements of prospectively collected lung and kidney cancer tumors. Kassandra-based TME reconstruction found the presence of PD1-positive CD8+ T cells strongly correlates with immunotherapy response and PD-L1 protein levels detected by immunohistochemistry, the established immunotherapy response biomarker, indicating that in the future RNA-seq-based cellular deconvolution could support oncology clinical decision-making.
“Our findings demonstrate the importance of cellular deconvolution based on RNA-seq to accurately understand the composition and activity of the tumor and the microenvironment to improve treatment outcomes,” said Nathan Fowler, MD, Chief Medical Officer at BostonGene. “We look forward to further developing the Kassandra algorithm and showing its clinical applicability in precision oncology.
BostonGene today announced the online publication of the manuscript, “Precise reconstruction of the tumor microenvironment using bulk RNA-seq and a unique machine learning algorithm trained on artificial transcriptomes” in Cancer Cell, a premier peer-reviewed scientific journal that publishes high impact results in cancer research and oncology. The study demonstrated the ability of the BostonGene-developed unique and robust machine learning (ML) algorithm named Kassandra to digitally reconstruct the tissue tumor microenvironment (TME) and blood cellular composition, identifying over 50 unique cell populations in total from RNA-seq derived from both archival and fresh tissues. These results indicate that cellular deconvolution can be utilized in future clinical applications to predict blood and tissue microenvironment composition, a critical factor in cancer pathogenesis, clinical outcome, and therapeutic resistance.
Kassandra was trained on a broad collection of > 9,400 tissue and blood-sorted cell RNA profiles incorporated into millions of artificial transcriptomes for this research study. To increase the stability and robustness of BostonGene’s decision-tree ML deconvolution algorithm, bioinformatic correction for technical and biological variability, aberrant cancer cell expression inclusion, and accurate quantification and normalization of transcript expression were applied. Results of the study showed that Kassandra recognizes 51 unique subpopulations and accurately deconvolves less well-characterized stromal and immune elements of prospectively collected lung and kidney cancer tumors. Kassandra-based TME reconstruction found the presence of PD1-positive CD8+ T cells strongly correlates with immunotherapy response and PD-L1 protein levels detected by immunohistochemistry, the established immunotherapy response biomarker, indicating that in the future RNA-seq-based cellular deconvolution could support oncology clinical decision-making.
“Our findings demonstrate the importance of cellular deconvolution based on RNA-seq to accurately understand the composition and activity of the tumor and the microenvironment to improve treatment outcomes,” said Nathan Fowler, MD, Chief Medical Officer at BostonGene. “We look forward to further developing the Kassandra algorithm and showing its clinical applicability in precision oncology.
About BostonGene Corporation
BostonGene powers an AI model of tumor and immune biology to deliver disease-level insights and enable precision decision-making across oncology and immune-mediated diseases, spanning drug development and clinical care. By integrating multimodal data, including genomic, transcriptomic, immune, and clinical signals, BostonGene generates biologically grounded intelligence to optimize patient selection, trial design, and therapeutic strategy. This creates a scalable AI decision layer that improves development outcomes and clinical management. BostonGene partners with leading biopharmaceutical organizations and academic institutions to accelerate the delivery of precision therapies while continuously expanding its capabilities across new diseases and complex biological systems. For more information, visit www.BostonGene.com.
Latest news
