Xueliang Wang

Watch-time prediction remains to be a key factor in reinforcing user engagement via video recommendations. It has become increasingly important given the ever-growing popularity of online videos. However, prediction of watch time not only depends on the match between the user and the video but is often mislead by the duration of the video itself. With the goal of improving watch time, recommendation is always biased towards videos with long duration. Models trained on this imbalanced data face the risk of bias amplification, which misguides platforms to over-recommend videos with long duration but overlook the underlying user interests. This paper presents the first work to study duration bias in watch-time prediction for video recommendation. We employ a causal graph illuminating that duration is a confounding factor that concurrently affects video exposure and watch-time prediction---the first effect on video causes the bias issue and should be eliminated, while the second effect on watch time originates from video intrinsic characteristics and should be preserved. To remove the undesired bias but leverage the natural effect, we propose a Duration-Deconfounded Quantile-based (D2Q) watch-time prediction framework, which allows for scalability to perform on industry production systems. Through extensive offline evaluation and live experiments, we showcase the effectiveness of this duration-deconfounding framework by significantly outperforming the state-of-the-art baselines. We have fully launched our approach on Kuaishou App, which has substantially improved real-time video consumption due to more accurate watch-time predictions.

BACKGROUND: Although human cancers have heterogeneous combinations of altered oncogenes, some crucial genes are universally dysregulated in most cancers. One such gene, FEAT (faint expression in normal tissues, aberrant overexpression in tumors), is uniformly overexpressed in a variety of human cancers and plays an important role in tumorigenesis by suppressing apoptosis. However, the precise molecular mechanism through which FEAT is upregulated during tumorigenesis remains largely unknown. METHODS: In this study, we used bioinformatic analyses to search for miRNAs that potentially target FEAT. We examined the expression of FEAT protein level by western blotting and miR-16 level by qRT-PCR assay. Cancer cell lines (A549, MCF-7 and Huh-7) with miR-16 upregulation and FEAT silencing were established and the effects on apoptosis of cancer cells in vitro were assessed. Luciferase reporter assay was also performed to investigate the interaction between miR-16 and FEAT. RESULTS: We identified a specific target site for miR-16 in the 3'-untranslated region (3'-UTR) of FEAT. Consistent with the bioinformatic analyses, we identified an inverse correlation between the miR-16 and FEAT protein levels in lung cancer, breast cancer, and hepatocellular cancer tissues. We then experimentally validated miR-16 as a direct regulator of FEAT using cell transfection and luciferase assays. Finally, we demonstrated that the repression of FEAT by miR-16 promoted the apoptosis of cancer cells. CONCLUSIONS: Our findings provide the first clues regarding the role of miR-16 as a tumor suppressor in cancer cells through the inhibition of FEAT translation.

MADS-box proteins, a well-conserved family of transcription factors in eukaryotic organisms, specifically regulate a wide range of cellular functions, including primary metabolism, cell cycle, and cell identity. However, little is known about roles of the MADS-box protein family in the fungal pathogen Sclerotinia sclerotiorum. In this research, the S. sclerotiorum MADS-box gene SsMADS was cloned; it encodes a protein that is highly similar to Mcm1 orthologs from Saccharomyces cerevisiae and other fungi, and includes a highly conserved DNA-binding domain. MADS is a member of the MADS box protein SRF (serum response factor) lineage. SsMADS function was investigated using RNA interference. Silenced strains were obtained using genetic transformation of the RNA interference vectors pS1-SsMADS and pSD-SsMADS. SsMADS expression levels in silenced strains were analyzed using RT-PCR. The results showed that SsMADS mRNA expression in these silenced strains was reduced to different degrees, and growth rate in these silenced strains was significantly decreased. Infecting tomato leaflets with silenced strains indicated that SsMADS was required for leaf pathogenesis in a susceptible host. Our results suggest that the MADS-box transcription factor SsMADS is involved in S. sclerotiorum growth and virulence.

The signature of the intrahepatic microbiome in multifocal hepatocellular carcinoma (HCC) and its association with genomic alterations remain elusive. In this study, we performed multiomics profiling of 242 HCC tumor nodules and 58 adjacent nontumor tissues from 58 patients with multifocal HCC, revealing heterogeneous microbial communities in multifocal HCC. The presence of bacteria in HCC nodules was confirmed by Gram stain, lipopolysaccharide, lipoteichoic acid staining, and transmission electron microscopy. Mutational profiling stratified patients into intrahepatic metastasis (IM)-HCC and multicentric occurrence (MO)-HCC. Bacterial communities differed between IM and MO nodules (P = 0.01). A nine-bacterium biomarker panel could distinguish IM nodules from MO nodules with an AUROC of 0.795. The epithelial-mesenchymal transition pathway was upregulated in IM nodules and correlated with IM-enriched bacteria. IM-enriched bacteria such as Enterococcus faecalis and Streptococcus anginosus promoted HCC cell migration, invasion, and tumor progression in orthotopic HCC mouse models by inducing an immunosuppressive microenvironment and epithelial-mesenchymal transition. Collectively, the intrahepatic microbiome contributes to the heterogeneity and pathogenesis of multifocal HCC. SIGNIFICANCE: We reveal intraindividual heterogeneous microbial communities among nodules from patients with multifocal HCC. IM-HCC-enriched bacteria promote tumor growth and influence the tumor microenvironment of HCC. Our work highlights the necessity of considering bacterial heterogeneity as biomarkers and targets for multifocal HCC therapeutic intervention. See related commentary by Dzutsev and Trinchieri, p. 1540.