Towards artificial general intelligence via a multimodal foundation model.

The fundamental goal of artificial intelligence (AI) is to mimic the core cognitive activities of human. Despite tremendous success in the AI research, most of existing methods have only single-cognitive ability. To overcome this limitation and take a solid step towards artificial general intelligence (AGI), we develop a foundation model pre-trained with huge multimodal data, which can be quickly adapted for various downstream cognitive tasks. To achieve this goal, we propose to pre-train our foundation model by self-supervised learning with weak semantic correlation data crawled from the Internet and show that promising results can be obtained on a wide range of downstream tasks. Particularly, with the developed model-interpretability tools, we demonstrate that strong imagination ability is now possessed by our foundation model. We believe that our work makes a transformative stride towards AGI, from our common practice of "weak or narrow AI" to that of "strong or generalized AI".

Class-Aware Sounding Objects Localization via Audiovisual Correspondence.

Audiovisual scenes are pervasive in our daily life. It is commonplace for humans to discriminatively localize different sounding objects but quite challenging for machines to achieve class-aware sounding objects localization without category annotations, i.e., localizing the sounding object and recognizing its category. To address this problem, we propose a two-stage step-by-step learning framework to localize and recognize sounding objects in complex audiovisual scenarios using only the correspondence between audio and vision. First, we propose to determine the sounding area via coarse-grained audiovisual correspondence in the single source cases. Then visual features in the sounding area are leveraged as candidate object representations to establish a category-representation object dictionary for expressive visual character extraction. We generate class-aware object localization maps in cocktail-party scenarios and use audiovisual correspondence to suppress silent areas by referring to this dictionary. Finally, we employ category-level audiovisual consistency as the supervision to achieve fine-grained audio and sounding object distribution alignment. Experiments on both realistic and synthesized videos show that our model is superior in localizing and recognizing objects as well as filtering out silent ones. We also transfer the learned audiovisual network into the unsupervised object detection task, obtaining reasonable performance.

MicroRNA-214 promotes chronic kidney disease by disrupting mitochondrial oxidative phosphorylation.

Mitochondria are critical in determining a cell's energy homeostasis and fate, and mitochondrial dysfunction has been implicated in the pathogenesis of chronic kidney disease (CKD). We sought to identify causative mitochondrial microRNAs. A microarray screen of kidney tissue from healthy mice identified 97 microRNAs that were enriched in the mitochondrial fraction. We focused on microRNA-214-3p (miR-214) because of a very high ratio of mitochondrial to cytoplasmic expression in the kidney compared to other organs. Tubular expression of miR-214 was more abundant in kidney tissue from patients with CKD than from healthy controls, and was positively correlated with the degree of proteinuria and kidney fibrosis. Expression of miR-214 was also increased in the kidney of mouse models of CKD induced by obstruction, ischemia/reperfusion, and albumin overload. Proximal tubule-specific deletion of miR-214 attenuated apoptosis, inflammation, fibrosis, and mitochondrial damage in these CKD models. Pharmacologic inhibition of miR-214 had a similar effect in the albumin overload model of CKD. In vitro, overexpressing miR-214 in proximal tubular cell lines induced apoptosis and disrupted mitochondrial oxidative phosphorylation, while miR-214 expression was upregulated in response to a variety of insults. The mitochondrial genes mt-Nd6 and mt-Nd4l were identified as the specific targets of miR-214 in the kidney. Together, these results demonstrate a pathogenic role of miR-214 in CKD through the disruption of mitochondrial oxidative phosphorylation, and suggest the potential for miR-214 to serve as a therapeutic target and diagnostic biomarker for CKD.

Rapid screening and identification of dominant B cell epitopes of HBV surface antigen by quantum dot-based fluorescence polarization assay.

A method for quickly screening and identifying dominant B cell epitopes was developed using hepatitis B virus (HBV) surface antigen as a target. Eleven amino acid fragments from HBV surface antigen were synthesized by 9-fluorenylmethoxy carbonyl solid-phase peptide synthesis strategy, and then CdTe quantum dots were used to label the N-terminals of all peptides. After optimizing the factors for fluorescence polarization (FP) immunoassay, the antigenicities of synthetic peptides were determined by analyzing the recognition and combination of peptides and standard antibody samples. The results of FP assays confirmed that 10 of 11 synthetic peptides have distinct antigenicities. In order to screen dominant antigenic peptides, the FP assays were carried out to investigate the antibodies against the 10 synthetic peptides of HBV surface antigen respectively in 159 samples of anti-HBV surface antigen-positive antiserum. The results showed that 3 of the 10 antigenic peptides may be immunodominant because the antibodies against them existed more widely among the samples and their antibody titers were higher than those of other peptides. Using three dominant antigenic peptides, 293 serum samples were detected for HBV infection by FP assays; the results showed that the antibody-positive ratio was 51.9% and the sensitivity and specificity were 84.3% and 98.2%, respectively. In conclusion, a quantum dot-based FP assay is a very simple, rapid, and convenient method for determining immunodominant antigenic peptides and has great potential in applications such as epitope mapping, vaccine designing, or clinical disease diagnosis in the future.

[Effect of p53 on the variation of renal tubular epithelial cells after kidney ischemia/reperfusion injury].

AIM: To observe the variation of renal tubular epithelial cells in p53(+/+) and p53(-/-) mice with young or old age at different time after kidney ischemia/reperfusion injury (IRI), and to investigate the contribution of p53 gene in the variation. METHODS: p53(+/+) and p53(-/-) male mice at age of 2 and 12 months were made ischemic by clamping left renal hila for 45 min. At 0, 1, 3 and 7 d, 1, 3 and 6 month after reflow, renal tissues were processed for morphometric observation and proliferating cell nuclear antigen(PCNA), apoptosis and senescence-associated beta-galactosidase (SA-beta-gal) analysis, using hematoxylin and eosin stain, immunohistochemistry, terminal deoxynucleotidyl transferase-mediated dUTP-biotin in situ nick-end labeling (TUNEL) and histochemical staining, respectively. RESULTS: Renal tubule necrosis was more severely in p53(-/-) mice and aged mice compared to p53(+/+) mice and young mice (P<0.05), respectively. Apoptotic cells in p53(+/+) mice increased obviously compared to p53(-/-) mice (P<0.05) at 7 d after IRI. In young wild-type mice, occasionally faint staining for SA-beta-gal activity began to appear at 1 month, and obviously significantly increased at 3 and 6 months after IRI (P<0.05), but in contralateral kidney at any time as mentioned above, and in the IRI kidneys in p53(-/-) mice at 1 and 3 months, there was almost no positive staining for SA-beta-gal; occasionally positive staining for SA-beta-gal was observed in the IRI kidney in p53(-/-) mice at 6 months after IRI. In p53 (-/-) and p53(+/+) aged mice, both kindeys had positive staining for SA-beta-gal activity at 0 d after IRI, but the level of the activity in p53(-/-) mice was much more lower than that in p53(+/+) mice (P<0.05), then the level of the activity decreased notably at 1 d in the IRI kidney (P<0.05). Positive stain of nuclear PCNA in p53(+/+) young mice had no statistical significance compared to p53(+/+) aged mice (P>0.05). But in p53(-/-) mice, significant positive staining for PCNA was tested, especially in young mice and in IRI kidneys (P>0.05). Correlation analysis between senescent and apoptotic cells in aged mice was made at 1 d after IRI, then striking negative correlation was found between both of them in p53(+/+) mice (r=-0.82, P<0.05), but no statistical correlation in p53(-/-) mice (r=0.26, P>0.05). CONCLUSION: IRI can accelerate renal tubular cell senescence and cellular death(both necrosis and apoptosis)after that. p53 gene may play an important role in the variation of tubular epithelial cells after kidney IRI.

[Effect of p21 on the changes in renal tubular epithelial cells after ischemia/reperfusion injury of kidney].

OBJECTIVE: To investigate the contribution of p21 gene in renal tubular epithelial cells in p21 (+/+) and p21 (-/-) mice of young and old ages at different times after kidney ischemia/reperfusion injury (IRI). METHODS: In p21 (+/+) and p21 (-/-) male mice at the ages of 2 and 12 months the kidneys were made ischemic by clamping the left renal artery for 45 minutes followed by declamping. On 0, 1, 3 and 7 days, 1, 3 and 6 months after reflow, renal tissue was processed for pathological study, determination of proliferating cell nuclear antigen (PCNA), apoptosis and senescence-associated beta-galactosidase (SA-beta-gal) analysis, using hematoxylin and eosin staining, immunohistochemistry, terminal deoxynucleotidyl transferase-mediated dUTP-biotin in situ nick-end labeling (TUNEL), and histochemical staining, respectively. RESULTS: Renal tubule necrosis and cell apoptosis were more severe in p21 (-/-) mice and old mice as compared with p21 (+/+) mice and young mice (both P<0.05), respectively. In young p21 (+/+) mice, occasionally faint staining for SA-beta-gal activity began to appear after 1 month, and significantly increased 3 and 6 months after IRI (P<0.05), but there was no positive staining for SA-beta-gal in the contralateral kidney or both kidneys in p21 (-/-) mice at any time. Another manner of the expression of SA-beta-gal was detected in aged p21 (+/+) mice, as both kidneys showed intensely positive staining for SA-beta-gal at 0 day after IRI, it then subsided notably on 1 day in the IRI kidney (P<0.05), but increased again at 3 months, though still less intense than the contralateral kidney, albeit more intense than the young mice at the same time (P<0.05). Three months after IRI, in both the IRI kidney and the contralateral kidney, positive staining for SA-beta-gal almost reached the same level. On the contrary, only occasional faint staining for SA-beta-gal activity was observed in aged p21 (-/-) mice at any time. No significant difference in positive staining of nuclear PCNA was found between in young and aged p21 (+/+) mice (P>0.05), although the numbers of positively stained nuclear PCNA were more in number in young mice than in aged mice. But in p21 (-/-) mice, significantly more cells were positively stained for PCNA, especially in young mice and in IRI kidneys (P<0.05). Correlation analysis between senescent and apoptotic cells in aged mice made at 1 day after IRI showed striking negative correlation between both of them [p21 (+/+) mice: r=-0.82; P<0.001; p21 (-/-) mice: r=-0.76, P<0.001]. CONCLUSION: IRI can promote the senescence process of normal tubular cells, and can accelerate death (necrosis and apoptosis) process of senescent tubular cells. p21 gene may play an important role in the senescence changes in tubular epithelial cells after kidney ischemia/reperfusion injury.

[Gly374Arg mutation in Fgfr3 causes achondroplasia in mice].

OBJECTIVE: To establish the mouse model of Gly374Arg mutation in fibroblast growth factor receptor 3(Fgfr3) and to analyze the phenotype of the mutant mice. METHODS: The double PCR was used to introduce Gly374Arg point mutation into mouse Fgfr3. The electroporation of embryonic stem(ES) cells was carried out with targeting vector. The targeted ES cells were screened by Positive-Negative Selection of G418 and Ganciclovir, and Southern blot. The correct targeted ES cells were microinjected into blastula. Finally, mutant mice were obtained by crossing between EIIa-Cre transgenic mice and mice carrying recombined mutant Fgfr3 allele. The mice were genotyped by PCR, and phenotype was observed by skeleton staining, histology, etc. RESULTS: Fgfr3-Gly374Arg mutant mice exhibited small size, short tail, macrocephaly and had dome-shaped heads, the epiphyseal growth plates of mutant mice were narrower, and the hypertrophic chondrocyte zone was also obviously decreased. Meanwhile, the majority of female mice were infertile, and the uterus, ovary and mammal gland in mutant female mice were also smaller and underdeveloped. CONCLUSION: The model of Fgfr3-Gly374Arg mutation causing achondroplasia in mice has been established successfully.