Single-Cell-Spatial Transcriptome Combined With "Generative Modeling" Deep Learning To Resolve The Spatio-Temporal Evolution Of Gene Regulatory Networks At The Invasive Front Of Bladder Cancer

Objective:Bladder cancer is one of the most common cancers of the urinary system.Worldwide,bladder cancer ranks 7th in incidence and 13 th in mortality among malignant tumors,and is extremely harmful to human beings.Invasive metastasis is the root cause of bladder cancer and even all malignant tumors becoming lethal,and there is a lack of drugs that effectively target metastasis in clinical practice.The invasion front is the most active region for tumor cell invasion and is the key to cause recurrent metastasis,and the alteration of its malignant phenotype is generally considered to be controlled by a gene regulatory network.However,molecular experiments and single-cell sequencing studies cannot measure spatial heterogeneity;pathology studies can neither assess transcriptional activity nor portray the full picture of transcriptional regulation;and none of the existing algorithms for the construction of single-cell gene regulatory networks can utilize spatial information."Assessing the spatial and temporal evolution of gene regulatory networks from the central to the invasive front of the tumor has become a common multidisciplinary problem.As bladder cancer is highly aggressive and prone to recurrence and progression,"gene regulatory networks reflecting the dynamic evolution of bladder cancer invasion front" is of great scientific value.The next common problem in the field of bioinformatics,computer science and medicine is that "the use of established gene regulatory networks is unclear and difficult to translate into the clinical setting".The purpose of this study is to explore the gene regulatory networks at the invasive front of bladder cancer and to construct a computational method that can reflect the spatial and temporal dynamics of gene regulatory networks in cancer cells.On the basis of the above,further research is aimed at solving the problem of "clinical translation" of gene regulatory networks and providing new intervention ideas to reverse the invasion and metastasis of bladder cancer.Methods: Using 8 cases of single-cell sequencing(sc RNA-seq)and 4 cases of spatial transcriptome sequencing from surgically resected bladder cancer specimens,we developed a general algorithm for dynamic gene regulatory networks of tumors incorporating spatio-temporal trajectories using generative deep learning technology;based on single-cell multi-omics data,we used the algorithm to construct a network to explore the dynamic evolution of gene regulatory networks at the invasive frontier of bladder cancer and to screen key regulatory nodes based on the network’s We will use the algorithm to construct a network to explore the dynamic evolution of the gene regulatory network at the forefront of bladder cancer invasion,and screen key regulatory nodes based on the "graph" properties of the network;we will use computer-aided drug screening,molecular dynamics simulation and other technologies,and apply the latest tools such as Alphafold2 to explore the drug potential of target genes and realize the "new use of old drugs".Results: In this study,we identified cell subpopulations located at the invasion front and described the spatio-temporal trajectories of the invasion subpopulations.Based on singlecell sequencing sc RNA-seq and spatial transcriptome data to achieve the localization of single cells in spatio-temporal trajectories,we constructed a dynamic gene regulatory network W incorporating spatio-temporal trajectories,screened and evaluated the key regulatory nodes of the network,HBXIP and C7ORF59,performed computer-assisted drug screening for target proteins Lamtor4 protein,Lamtor5 protein and Lamtor4-Lamtor5 protein complexes,and dihydroergotamine was screened for possible reversal of bladder cancer invasion and metastasis.Conclusion: This general algorithm can construct a dynamic gene regulatory network (2 incorporating spatio-temporal trajectories to achieve dynamic evolutionary inference of the gene regulatory network of bladder cancer invasion front,target genes HBXIP and C7ORF59 as possible drug targets,and dihydroergotamine screening as available drugs that may reverse bladder cancer invasion metastasis.