Context Nonfunctioning pituitary adenomas (NFPAs) are the major cause of hypopituitarism and infertility. However, the pathogenesis of NFPAs remains largely unknown. Previous studies have demonstrated the crucial role of miRNAs in the progression of pituitary adenomas. Increasing evidence has indicated that circular RNAs (circRNAs) might mediate miRNA transcriptional activity, providing new insights to study the pathogenesis of NFPAs. Objectives To explore the regulation and activity of the circRNA–miRNA–mRNA axis in the tumorigenesis of NFPAs. Design The function of miR-145-5p in NFPAs was investigated invitro and invivo. The mechanical details were explored and potential targets of miR-145-5p were identified. Finally, miR-145-5p–associated circRNAs were functionally recognized and confirmed. Results miR-145-5p was markedly decreased in NFPA samples and correlated negatively with NFPA invasiveness. Overexpression of miR-145-5p suppressed NFPA cell proliferation and invasiveness and promoted apoptosis. Further results confirmed that translationally controlled tumor protein (TPT1) is a target of miR-145-5p and mediated the effect of miR-145-5p. TPT1 and its downstream factors Mcl-1 and Bcl-xL were downregulated, and Bax was upregulated by miR-145-5p. Moreover, circOMA1 (hsa_circRNA_0002316) was demonstrated to sponge miR-145-5p, whose suppression on NFPA cells was abrogated by circOMA1 overexpression. circOMA1 silencing exhibited a similar inhibitory effect with miR-145-5p overexpression by downregulating TPT1. We found that circOMA1 could further upregulate Mcl-1 and Bcl-xL and downregulate Bax. Conclusions circOMA1 promotes NFPA progression by acting as the sponge of tumor suppressor miR-145-5p to regulate the TPT1 signaling pathway, revealing a therapeutic target in preventing the tumorigenesis of NFPAs.

Background: A growing number of evidences suggest that TMZ applications can generate impressive benefits for APT and PC patients. However, the definite role of TMZ for individuals remains unclarified due to the variation between studies. And the predictive factors to alter its efficacy remain debatable.Objective: To evaluate the long-term effectiveness and safety profile of TMZ in the treatment of pituitary malignancies, and delineate the predictors during its clinical employment.Results: A literature retrieval was conducted from online databases for studies published up to December 31, 2020. Twenty one studies involving 429 patients were identified. TMZ exhibited 41% radiological overall response rate (rORR). The biochemical response rate was determinate in 53% of the functioning subset. Two-year and 4-year survival rate were 79 and 61%, respectively. TMZ prolonged the median PFS and OS as 20.18 and 40.24 months. TMZ-related adverse events occurred in 19% of patients. Regarding predictors of TMZ response, rORR was dramatically improved in patients with low/intermediate MGMT expression than those with high-MGMT (>50%) (p < 0.001). The benefit of TMZ varied according to functioning subtype of patients, with greater antitumor activities in functioning subgroups and fewer activities in non-functioning sets (p < 0.001). Notably, the concomitant therapy of radiotherapy and TMZ significantly increased the rORR (p = 0.007).Conclusion: TMZ elicits clinical benefits with moderate adverse events in APT and PC patients. MGMT expression and clinical subtype of secreting function might be vital predictors of TMZ efficacy. In the future, the combination of radiotherapy with TMZ may further improve the clinical outcomes than TMZ monotherapy.

In recent years, the Mixture-of-Experts (MoE) technique has gained widespread popularity as a means to scale pre-trained models to exceptionally large sizes. Dynamic activation of experts allows for conditional computation, increasing the number of parameters of neural networks, which is critical for absorbing the vast amounts of knowledge available in many deep learning areas. However, despite the existing system and algorithm optimizations, there are significant challenges to be tackled when it comes to the inefficiencies of communication and memory consumption. In this paper, we present the design and implementation of MPMoE, a high-performance library that accelerates MoE training with adaptive and memory-efficient pipeline parallelism. Inspired by that the MoE training procedure can be divided into multiple independent sub-stages. We design a pipeline parallelism method for reducing communication latency by overlapping with computation operations. Further, we analyze the memory footprint breakdown of MoE training and identify that activations and temporary buffers are the primary contributors to the overall memory footprint. Toward memory efficiency, we propose memory reuse strategies to reduce memory requirements by eliminating memory redundancies. Finally, to optimize pipeline granularity and memory reuse strategies jointly, we propose a profile-based algorithm and a performance model to determine the configurations of MPMoE at runtime. We implement MPMoE upon PyTorch and evaluate it with common MoE models in two physical clusters, including 64 NVIDIA A100 GPU cards and 16 NVIDIA V100 GPU cards. Compared with the state-of-art approach, MPMoE achieves up to 2.3× speedup while reducing more than 30% memory footprint for training large models.