The privacy-explainability trade-off: unraveling the impacts of differential privacy and federated learning on attribution methods.

Since the advent of deep learning (DL), the field has witnessed a continuous stream of innovations. However, the translation of these advancements into practical applications has not kept pace, particularly in safety-critical domains where artificial intelligence (AI) must meet stringent regulatory and ethical standards. This is underscored by the ongoing research in eXplainable AI (XAI) and privacy-preserving machine learning (PPML), which seek to address some limitations associated with these opaque and data-intensive models. Despite brisk research activity in both fields, little attention has been paid to their interaction. This work is the first to thoroughly investigate the effects of privacy-preserving techniques on explanations generated by common XAI methods for DL models. A detailed experimental analysis is conducted to quantify the impact of private training on the explanations provided by DL models, applied to six image datasets and five time series datasets across various domains. The analysis comprises three privacy techniques, nine XAI methods, and seven model architectures. The findings suggest non-negligible changes in explanations through the implementation of privacy measures. Apart from reporting individual effects of PPML on XAI, the paper gives clear recommendations for the choice of techniques in real applications. By unveiling the interdependencies of these pivotal technologies, this research marks an initial step toward resolving the challenges that hinder the deployment of AI in safety-critical settings.

Integrated analysis of ALK higher expression in human cancer and downregulation in LUAD using RNA molecular scissors

PurposeAnaplastic lymphoma kinase (ALK) is an endorsed molecular target in ALK-rearranged carcinomas, including lung adenocarcinoma. However, the clinical advantage of targeting ALK using druggable inhibitors is almost universally restricted by the development of drug resistance. Therefore, a strategy for combating ALK overexpression remains paramount for ALK-driven cancer.MethodsWe systemically analyzed the overexpression pattern of ALK and its clinical consequences, genetic alterations, and their significance in cancer hallmark genes, and correlation using integrated multidimensional approaches. The LwCas13a RNA molecular scissors was used to downregulate ALK-rearrangement by leveraging two target guide RNAs in lung adenocarcinoma (LUAD) cells. Immunocytochemistry, immunoblotting, and MTT assays were conducted to validate the downregulation.ResultsWe found elevated levels of ALK in several malignancies, including LUAD, than in normal tissues. Higher expression of ALK was significantly associated with worse or shorter survival than patients with lower expression. We identified numerous genetic alterations in ALK, which potentially alter the cancer hallmark genes, including STAT1 and CTSL, in patients with LUAD. Next, we observed that the LwCas13a molecular scissors robustly downregulated both phosphorylated and total ALK chimera protein expression in LUAD cells compared to the control. Furthermore, we found that downregulation of ALK chimera protein substantially inhibited cell viability and induced cell death, including apoptosis.ConclusionOur findings suggest a basis for ALK as a prognostic biomarker and the LwCas13a molecular scissors successfully downregulated the onco-driver ALK-rearrangement protein, which will potentially pave the way toward the development of novel therapeutic strategies for ALK-driven cancer. (AU)