Subject-driven text-to-image generation aims to generate customized images of the given subject based on the text descriptions, which has drawn increasing attention. Existing methods mainly resort to finetuning a pretrained generative model, where the identity-relevant information (e.g., the boy) and the identity-irrelevant information (e.g., the background or the pose of the boy) are entangled in the latent embedding space. However, the highly entangled latent embedding may lead to the failure of subject-driven text-to-image generation as follows: (i) the identity-irrelevant information hidden in the entangled embedding may dominate the generation process, resulting in the generated images heavily dependent on the irrelevant information while ignoring the given text descriptions; (ii) the identity-relevant information carried in the entangled embedding can not be appropriately preserved, resulting in identity change of the subject in the generated images. To tackle the problems, we propose DisenBooth, an identity-preserving disentangled tuning framework for subject-driven text-to-image generation. Specifically, DisenBooth finetunes the pretrained diffusion model in the denoising process. Different from previous works that utilize an entangled embedding to denoise each image, DisenBooth instead utilizes disentangled embeddings to respectively preserve the subject identity and capture the identity-irrelevant information. We further design the novel weak denoising and contrastive embedding auxiliary tuning objectives to achieve the disentanglement. Extensive experiments show that our proposed DisenBooth framework outperforms baseline models for subject-driven text-to-image generation with the identity-preserved embedding. Additionally, by combining the identity-preserved embedding and identity-irrelevant embedding, DisenBooth demonstrates more generation flexibility and controllability.

Graph Neural Architecture Search (GNAS) has achieved superior performance on various graph-structured tasks. However, existing GNAS studies overlook the applications of GNAS in resource-constraint scenarios. This paper proposes to design a joint graph data and architecture mechanism, which identifies important sub-architectures via the valuable graph data. To search for optimal lightweight Graph Neural Networks (GNNs), we propose a Lightweight Graph Neural Architecture Search with Graph SparsIfication and Network Pruning (GASSIP) method. In particular, GASSIP comprises an operation-pruned architecture search module to enable efficient lightweight GNN search. Meanwhile, we design a novel curriculum graph data sparsification module with an architecture-aware edge-removing difficulty measurement to help select optimal sub-architectures. With the aid of two differentiable masks, we iteratively optimize these two modules to efficiently search for the optimal lightweight architecture. Extensive experiments on five benchmarks demonstrate the effectiveness of GASSIP. Particularly, our method achieves on-par or even higher node classification performance with half or fewer model parameters of searched GNNs and a sparser graph.

BACKGROUND: Being among the most common malignancies worldwide, hepatocellular carcinoma (HCC) accounting for the third cause of cancer mortality. The regulation of cell death is the most crucial step in tumor progression and has become a crucial target for nearly all therapeutic options. Cuproptosis, a copper-induced cell death, was recently reported in Science. However, its primary function in carcinogenesis is still unclear. METHODS: Cuproptosis-related lncRNAs significantly associated with overall survival (OS) were screened by stepwise univariate Cox regression. The signature of cuproptosis-related lncRNAs for HCC prognosis was constructed by the LASSO algorithm and multivariate Cox regression. Further Kaplan-Meier analysis, proportional hazards model, and ROC analysis were performed. Functional annotation was performed using gene set enrichment analysis (GSEA). The relationship between prognostic cuproptosis-related lncRNAs and HCC prognosis was further explored by GEPIA( http://gepia.cancer-pku.cn/ ) online analysis tool. Finally, we used the ESTIMATE and XCELL algorithms to estimate stromal and immune cells in tumor tissue and cast each sample to infer the underlying mechanism of cuproptosis-related lncRNAs in the tumor immune microenvironment (TIME) of HCC patients. RESULTS: Four cuproptosis-related lncRNAs were used to construct a prognostic lncRNA signature, which was an independent factor in predicting OS in HCC patients. Kaplan-Meier curves showed significant differences in survival rates between risk subgroups (p = 0.002). At the same time, we found that the expression levels of most immune checkpoint genes increased with increasing risk scores. Tumorigenesis and immunological-related pathways were primarily enhanced in the high-risk group, as determined by GSEA. The results of drug sensitivity analysis showed that compared with patients in the high-risk group, the IC50 values of erlotinib and lapatinib were lower in patients in the low-risk group, while the opposite was true for sunitinib, paclitaxel, gemcitabine, and imatinib. We also found that elevated AL133243.2 expression was significantly associated with worse OS and disease-free survival (DFS), more advanced T stage and higher tumor grade, and reduced immune cell infiltration, suggesting that HCC patients with low AL133243.2 expression in tumor tissues may have a better response to immunotherapy. CONCLUSION: Collectively, the cuproptosis-associated lncRNA signature can serve as an independent predictor to guide individual treatment strategies. Furthermore, AL133243.2 is a promising marker for predicting immunotherapy response in HCC patients. This data may facilitate further exploration of more effective immunotherapy strategies for HCC.