Comparing the Pretrained Models of Source Code by Re-pretraining Under a Unified Setup.

Recent years have seen the successful application of large pretrained models of source code (CodePTMs) to code representation learning, which have taken the field of software engineering (SE) from task-specific solutions to task-agnostic generic models. By the remarkable results, CodePTMs are seen as a promising direction in both academia and industry. While a number of CodePTMs have been proposed, they are often not directly comparable because they differ in experimental setups such as pretraining dataset, model size, evaluation tasks, and datasets. In this article, we first review the experimental setup used in previous work and propose a standardized setup to facilitate fair comparisons among CodePTMs to explore the impacts of their pretraining tasks. Then, under the standardized setup, we re-pretrain CodePTMs using the same model architecture, input modalities, and pretraining tasks, as they declared and fine-tune each model on each evaluation SE task for evaluating. Finally, we present the experimental results and make a comprehensive discussion on the relative strength and weakness of different pretraining tasks with respect to each SE task. We hope our view can inspire and advance the future study of more powerful CodePTMs.

KDM2B mediates the Wnt/ß-catenin pathway through transcriptional activation of PKMYT1 via microRNA-let-7b-5p/EZH2 to affect the development of non-small cell lung cancer.

The significance of KDM2B in oncogenesis has been appreciated, but the mechanism behind is incompletely understood. In this work, we addressed its effects on the progression of non-small cell lung cancer (NSCLC). Overexpression of KDM2B was linked to dismal prognoses of NSCLC patients. Based on the expression levels of KDM2B in a panel of NSCLC cell lines, A549, showing lower level of expression, and SK-MES-1, showing higher levels of expression, were selected as model systems to evaluate the effect of KDM2B overexpression and KDM2B silencing, respectively. Knockdown of KDM2B hampered NSCLC cell proliferation, invasion, as well as migration, while enhanced apoptosis. Additionally, KDM2B repressed the expression of microRNA (miR)-let-7b-5p through demethylation modification of H3K36me2, thereby promoting the expression of zester homolog 2 (EZH2), the target gene of let-7b-5p in NSCLC. Moreover, EZH2 transcriptionally induced the expression of PKMYT1 to activate the Wnt/ß-catenin pathway. Sh-EZH2 and sh-PKMYT1 neutralized the supporting effects of KDM2B on cell proliferation, invasion and migration. Additionally, deletion of KDM2B reduced the xenograft volumes in nude mice. In conclusion, KDM2B induces the EZH2/PKMYT1/Wnt/ß-catenin axis by inhibiting the let-7b-5p expression, which promotes NSCLC growth. More investigations are essential to determine the oncogenic role of KDM2B in NSCLC.

Implementation of artificial intelligence in the histological assessment of pulmonary subsolid nodules.

BACKGROUND: Clinical management of subsolid nodules (SSNs) is defined by the suspicion of tumor invasiveness. We sought to develop an artificial intelligent (AI) algorithm for invasiveness assessment of lung adenocarcinoma manifesting as radiological SSNs. We investigated the performance of this algorithm in classification of SSNs related to invasiveness. METHODS: A retrospective chest computed tomography (CT) dataset of 1,589 SSNs was constructed to develop (85%) and internally test (15%) the proposed AI diagnostic tool, SSNet. Diagnostic performance was evaluated in the hold-out test set and was further tested in an external cohort of 102 SSNs. Three thoracic surgeons and three radiologists were required to evaluate the invasiveness of SSNs on both test datasets to investigate the clinical utility of the proposed SSNet. RESULTS: In the differentiation of invasive adenocarcinoma (IA), SSNet achieved a similar area under the curve [AUC; 0.914, 95% confidence interval (CI): 0.813-0.987] with that of the 6 doctors (0.900, 95% CI: 0.867-0.922). When interpreting with the assistance of SSNet, the sensitivity of junior doctors, specificity of senior doctor, and their accuracy were significantly improved. In the external test, SSNet (AUC: 0.949, 95% CI: 0.884-1.000) achieved a better AUC than doctors (AUC: 0.883, 95% CI: 0.826-0.939) whose AUC increased (AUC: 0.908, 95% CI: 0.847-0.982) with SSNet assistance. In the histological subtype classifications, SSNet achieved better performance than practicing doctors. The AUCs of doctors were significantly improved with the assistance of SSNet in both 4-category and 3-category classifications to 0.836 (95% CI: 0.811-0.862) and 0.852 (95% CI: 0.825-0.882), respectively. CONCLUSIONS: The AI diagnostic system achieved non-inferior performance to doctors, and will potentially improve diagnostic performance and efficiency in SSN evaluation.

Supplement CT-Guided Microcoil Placement for Localising Ground-glass Opacity (GGO) Lesions at "Blind Areas" of the Conventional Hook-Wire Technique.

BACKGROUND: In the conventional hook-wire technique of pulmonary nodular localisations there are several "blind areas", including the mediastinum-vicinity region, interlobar fissure-neighbouring areas and scapulae-shadowed areas. The present study aims to summarise the experiences of CT-guided microcoil placement as an alternative method for localising pulmonary ground-glass opacity (GGO) lesions before thoracoscopic wedge resections. METHODS: Sixteen GGO lesions at "blind areas" in 16 patients were localised with platinum-fibered microcoils under CT assistance before undergoing video-assisted thoracoscopic surgical resections. Information regarding coil placement, operations and complications was recorded. RESULTS: Of all lesions, 1 was in the mediastinum-vicinity region, 8 were covered by the scapulae, and 7 were close to interlobar fissures (3 horizontal fissures, 4 oblique fissures). All 16 (100%) lesions had been successfully marked with microcoils. No major complications of the puncture procedure occurred; there were only minor pneumothorax (n=2) and haemoptysis (n=1) complications, which required no intervention before operations. All GGO lesions and microcoils were successfully removed by initial wedge resections. Of the 16 lesions in "blind areas", 8 were adenocarcinoma in situ (AIS), 4 were minimally invasive adenocarcinoma (MIA), 3 were atypical adenomatous hyperplasia (AAH), and 1 was interstitial fibrous tissue proliferation. No major complications occurred postoperatively. CONCLUSIONS: For the "blind areas" of the hook-wire technique, CT-guided microcoil placement is an effective method of marking GGO lesions that makes thoracoscopic wedge resection easier.