Few-shot image generation with reverse contrastive learning.

Generative models, such as Generative Adversarial Networks (GANs), have recently shown remarkable capabilities in various generation tasks. However, the success of these models heavily depends on the availability of a large-scale training dataset. When the size of the training dataset is limited, the quality and diversity of the generated results suffer from severe degradation. In this paper, we propose a novel approach, Reverse Contrastive Learning (RCL), to address the problem of high-quality and diverse image generation under few-shot settings. The success of RCL benefits from a two-sided, powerful regularization. Our proposed regularization is designed based on the correlation between generated samples, which can effectively utilize the latent feature information between different levels of samples. It does not require any auxiliary information or augmentation techniques. A series of qualitative and quantitative results show that our proposed method is superior to the existing State-Of-The-Art (SOTA) methods under the few-shot setting and is still competitive under the low-shot setting, showcasing the effectiveness of RCL. Code will be released upon acceptance at https://github.com/gouayao/RCL.

Immune desert in MMR-deficient tumors predicts poor responsiveness of immune checkpoint inhibition.

Background: Although many efforts have been devoted to identify biomarkers to predict the responsiveness of immune checkpoint inhibitors, including expression of programmed death-ligand 1 (PD-L1) and major histocompatibility complex (MHC) I, microsatellite instability (MSI), mismatch repair (MMR) defect, tumor mutation burden (TMB), tertiary lymphoid structures (TLSs), and several transcriptional signatures, the sensitivity of these indicators remains to be further improved. Materials and methods: Here, we integrated T-cell spatial distribution and intratumor transcriptional signals in predicting the response to immune checkpoint therapy in MMR-deficient tumors including tumors of Lynch syndrome (LS). Results: In both cohorts, MMR-deficient tumors displayed personalized tumor immune signatures, including inflamed, immune excluded, and immune desert, which were not only individual-specific but also organ-specific. Furthermore, the immune desert tumor exhibited a more malignant phenotype characterized by low differentiation adenocarcinoma, larger tumor sizes, and higher metastasis rate. Moreover, the tumor immune signatures associated with distinct populations of infiltrating immune cells were comparable to TLSs and more sensitive than transcriptional signature gene expression profiles (GEPs) in immunotherapy prediction. Surprisingly, the tumor immune signatures might arise from the somatic mutations. Notably, patients with MMR deficiency had benefited from the typing of immune signatures and later immune checkpoint inhibition. Conclusion: Our findings suggest that compared to PD-L1 expression, MMR, TMB, and GEPs, characterization of the tumor immune signatures in MMR-deficient tumors improves the efficiency of predicting the responsiveness of immune checkpoint inhibition.

SRF-derived miR210 and miR30c both repress beating cardiomyocyte formation in the differentiation system of embryoid body.

Serum response factor (SRF) cooperates with various co-factors to manage the specification of diverse cell lineages during heart development. Many microRNAs mediate the function of SRF in this process. However, how are miR210 and miR30c involved in the decision of cardiac cell fates remains to be explored. In this study, we found that SRF directly controlled the cardiac expression of miR210. Both miR210 and miR30c blocked the formation of beating cardiomyocyte during embryoid body (EB) differentiation, a cellular model widely used for studying cardiogenesis. Both of anticipated microRNA targets and differentially expressed genes in day8 EBs were systematically determined and enriched with gene ontology (GO), Kyoto encyclopedia of genes and genomes (KEGG) and Reactome. Functional enrichments of prediction microRNA targets and down-regulated genes in day8 EBs of miR210 suggested the importance of PI3K-Akt signal and ETS2 in miR210 inhibition of cardiomyocyte differentiation. Similar analyses revealed that miR30c repressed both developmental progress and the adrenergic signaling in cardiomyocytes during the differentiation of EBs. Taken together, SRF directs the expression of miR210 and miR30c, and they repress cardiac development via inhibiting the differentiation of cardiac muscle cell lineage as well as the cell proliferation. Through the regulation of specific microRNAs, the complication of SRF's function in heart development is emphasized.