Share
Decoding Complexity: Insights into AI Explainability, Immune Repertoires, and Cancer Resistance
Rohita Biswas
Affiliation: Journal of Science, Humanities and Arts (JOSHA), Freiburg im Breisgau, Germany
Keywords: AI Explainability; Vision Transformers; CAR T-cell Resistance; Immune Repertoire Profiling; Machine Learning
Categories: CRIION
DOI: 10.17160/josha.13.1.1082
Languages: English
JOSHA presents a synthesis of recent research from the Collaborative Research Institute Intelligent Oncology (CRIION), highlighting advances in AI explainability, immune profiling, and cancer therapy resistance. The first study introduces Salvage, a Shapley-distribution–based method that improves interpretability in Vision Transformers through guided sampling. The second explores TRBC1/TRBC2 oligoclonality in T-cell lymphomas, revealing how tumor heterogeneity can drive primary resistance to TRBC-directed CAR T cell therapies. The third examines adenosine deaminase 2 deficiency, uncovering skewed T- and B-cell receptor repertoires that enable accurate patient identification via machine learning. Collectively, these works decode complexity across disciplines, offering new strategies to interpret AI models, map immune landscapes, and address therapeutic challenges, thereby advancing both research insight and clinical application in intelligent oncology and immunology.