Biology subjects

Hu, Z.

Publications and source records attributed to Hu, Z..

At least 307 records · Page 17Linked to original sources

bioRxiv2020

Combination of favorable gene haplotypes plays an important role in rice hybrid grain quality traits based on genome-wide association analysis

Yield level and grain quality determine the commercial potential of rice (Oryza sativa L.) varieties. Mining and using genes that control important rice grain quality characteristics are major tasks for plant breeders. Here, a genome-wide association study was conducted to determine the genetic bases of 12 rice grain quality traits in 113 varieties and 565 testcross hybrids. A total of 56 significant SNPs were associated with 9 of the traits in variety phenotypic, general combining ability, testcross hybrid phenotypic and mid-parental heterotic datasets, from which 45 novel loci were identified. The cumulative effects of superior alleles or favorable haplotypes of genes closest to significant quality trait-associated loci were found in the four datasets. Additionally, the favorable gene haplotypes performed better than those of superior alleles in the variety and testcross hybrid datasets. Pyramiding the favorable haplotypes of five cloned rice grain quality genes resulted in a very low amylose content and high yield in the latter. These testcross hybrids had rice grain qualities similar to their parental lines but with much higher yields. The amylose content, grain width and grain length predictions for potential hybrids among the 113 varieties using genomic selection based on the BayesB method revealed a performance trend similar to those the testcross hybrids in our study. Thus, the selection of combination of favorable rice quality-related gene haplotypes is recommended to breed hybrids with high yields and elite grain qualities.

genetics↗

bioRxiv2020

An extracellular protein regulates patched-related/DAF-6-mediated sensory compartment formation in C. elegans

Coordination of neurite extension with surrounding glia development is critical for neuronal function, but the underlying molecular mechanisms remain poorly understood. Through a genome-wide mutagenesis screen in C. elegans, we identified dyf-4 and daf-6 as two mutants sharing similar defects in dendrite extension. DAF-6 encodes a glia-specific patched-related membrane protein that plays vital roles in glial morphogenesis. We cloned dyf-4 and found that DYF-4 encodes a glia-secreted extracellular protein. Intriguingly, DYF-4 colocalizes with DAF-6 along the glial channel. Further investigations revealed that DYF-4 directly interacts with DAF-6 and regulates its proper membrane localization. Notably, reported glial suppressors of daf-6 could also restore dendrite elongation and ciliogenesis in both dyf-4 and daf-6 mutants. Collectively, our data suggest that secreted DYF-4 likely acts as a novel ligand/regulator for the patched-related receptor DAF-6 which promotes the proper formation of the glial channel and indirectly affects neurite extension and ciliogenesis.

developmental biology↗

bioRxiv2020

Networks of Splice Factor Regulation by Unproductive Splicing Coupled With Nonsense Mediated mRNA Decay

BackgroundNonsense mediated mRNA decay (NMD) is an RNA surveillance pathway that degrades aberrant transcripts harboring premature termination codons. This pathway, in conjunction with alternative splicing, regulates gene expression post-transcriptionally. Nearly all serine and arginine-rich (SR) proteins and many heterogeneous nuclear ribonucleoproteins (hnRNPs) produce isoforms that can be degraded by the NMD pathway. Many splicing factors have been reported to be regulated via alternative splicing coupled to NMD. However, it is still uncharacterized that to what extent NMD contributes to the regulation of splicing factors. ResultsHere, we characterized a regulatory network of splicing factors through alternative splicing coupled to NMD. Based upon an extensive literature search, we first assembled a network that encompasses the current knowledge of splice factors repressing or activating the expression of other splicing factors through alternative splicing coupled to NMD. This regulatory network is limited, including just a handful of well-studied splicing factors. To gain a more global and less biased overview, we examined the splicing factor-mRNA interactions from public crosslinking-immunoprecipitation (CLIP)-seq data, which provides information about protein-RNA interactions. A network view of these interactions reveals extensive binding among splicing regulators. We also found that splicing factors bind more frequently to transcripts of other splicing factors than to other genes. In addition, many splicing factors are targets of NMD, and might be regulated via alternative splicing coupled to NMD, which is demonstrated by the significant overlap between the experimental network and eCLIP-network. We found that hierarchy of the splicing-factor interaction network differs from the hierarchy observed for transcription factors. ConclusionThe extensive interaction between splicing factors and transcripts of other splicing factors suggests that the potential regulation via alternative splicing coupled with NMD is widespread. The splicing factor regulation is fundamentally different from that of transcription factors.

molecular biology↗

bioRxiv2020

StrVCTVRE: A supervised learning method to predict the pathogenicity of human structural variants.

BackgroundWhole genome sequencing resolves many clinical cases where standard diagnostic methods have failed. However, at least half of these cases remain unresolved after whole genome sequencing. Structural variants (SVs; genomic variants larger than 50 base pairs) of uncertain significance are the genetic cause of a portion of these unresolved cases. As sequencing methods using long or linked reads become more accessible and structural variant detection algorithms improve, clinicians and researchers are gaining access to thousands of reliable SVs of unknown disease relevance. Methods to predict the pathogenicity of these SVs are required to realize the full diagnostic potential of long-read sequencing. ResultsTo address this emerging need, we developed StrVCTVRE to distinguish pathogenic SVs from benign SVs that overlap exons. In a random forest classifier, we integrated features that capture gene importance, coding region, conservation, expression, and exon structure. We found that features such as expression and conservation are important but are absent from SV classification guidelines. We leveraged multiple resources to construct a size-matched training set of rare, putatively benign and pathogenic SVs. StrVCTVRE performs accurately across a wide SV size range on independent test sets, which will allow clinicians and researchers to eliminate about half of SVs from consideration while retaining a 90% sensitivity. ConclusionsWe anticipate clinicians and researchers will use StrVCTVRE to prioritize SVs in patients where no SV is immediately compelling, empowering deeper investigation into novel SVs to resolve cases and understand new mechanisms of disease. StrVCTVRE runs rapidly and is available at https://compbio.berkeley.edu/proj/strvctvre/.

genomics↗

bioRxiv2020

Isolating multiple formats of human monoclonal neutralizing antibodies against SARS-CoV-2 by in vitro site-directed antibody screening

Neutralizing antibody is one of the most effective interventions for acute pathogenic infection. Currently, over three million people have been identified for SARS-CoV-2 infection but SARS-CoV-2-specific vaccines and neutralizing antibodies are still lacking. SARS-CoV-2 infects host cells by interacting with angiotensin converting enzyme-2 (ACE2) via the S1 receptor-binding domain (RBD) of its surface spike glycoprotein. Therefore, blocking the interaction of SARS-CoV-2-RBD and ACE2 by antibody would cause a directly neutralizing effect against virus. In the current study, we selected the ACE2 interface of SARS-CoV-2-RBD as the targeting epitope for neutralizing antibody screening. We performed site-directed screening by phage display and finally obtained one IgG antibody (4A3) and several domain antibodies. Among them, 4A3 and three domain antibodies (4A12, 4D5, and 4A10) were identified to act as neutralizing antibodies due to their capabilities to block the interaction between SARS-CoV-2-RBD and ACE2-positive cells. The domain antibody 4A12 was predicted to have the best accessibility to all three ACE2-interfaces on the spike homotrimer. Pseudovirus and authentic SARS-CoV-2 neutralization assays showed that all four antibodies could potently protect host cells from virus infection. Overall, we isolated multiple formats of SARS-CoV-2-neutralizing antibodies via site-directed antibody screening, which could be promising candidate drugs for the prevention and treatment of COVID-19.

immunology↗

bioRxiv2020

Transcriptome Comparison of organ of Corti in mice, chick and zebrafish

DisclaimerThis manuscript has been withdrawn by the authors as it was submitted and made public without the full consent of all the authors. The authors would like to redesign and perform additional experimental work. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.Competing Interest StatementThe authors have declared no competing interest.

bioinformatics↗

bioRxiv2020

CRISPR/Cpf1 mediated genome editing enhances Bombyx mori resistance to BmNPV

CRISPR/Cas12a (Cpf1) is a single RNA-guided endonuclease that provides new opportunities for targeted genome engineering through the CRISPR/Cas9 system. Only AsCpf1 have been developed for insect genome editing, and the novel Cas12a orthologs nucleases and editing efficiency require more study in insect. We compared three Cas12a orthologs nucleases, AsCpf1, FnCpf1, and LbCpf1, for their editing efficiencies and antiviral abilities in vitro. The three Cpf1 efficiently edited the BmNPV genome and inhibited BmNPV replication in BmN-SWU1 cells. The antiviral ability of the FnCpf1 system was more efficient than the SpCas9 system after infection by BmNPV. We created FnCpf1xgIE1 and SpCas9xsgIE1 transgenic hybrid lines and evaluated the gene editing efficiency of different systems at the same target site. We improved the antiviral ability using the FnCpf1 system in transgenic silkworm. This study demonstrated use of the CRISPR/Cpf1 system to achieve high editing efficiencies in the silkworm, and illustrates the use of this technology for increasing disease resistance. Author SummaryGenome editing is a powerful tool that has been widely used in gene function, gene therapy, pest control, and disease-resistant engineering in most parts of pathogens research. Since the establishment of CRISPR/Cas9, powerful strategies for antiviral therapy of transgenic silkworm have emerged. Nevertheless, there is still room to expand the scope of genome editing tool for further application to improve antiviral research. Here, we demonstrate that three Cpf1 endonuclease can be used efficiency editing BmNPV genome in vitro and in vivo for the first time. More importantly, this Cpf1 system could improve the resistance of transgenic silkworms to BmNPV compare with Cas9 system, and no significant cocoons difference was observed between transgenic lines infected with BmNPV and control. These broaden the range of application of CRISPR for novel genome editing methods in silkworm and also enable sheds light on antiviral therapy.

microbiology↗

bioRxiv2020

Accurate Diagnosis of Colorectal Cancer Based on Histopathology Images Using Artificial Intelligence

BackgroundAccurate and robust pathological image analysis for colorectal cancer (CRC) diagnosis is time-consuming and knowledge-intensive, but is essential for CRC patients treatment. The current heavy workload of pathologists in clinics/hospitals may easily lead to unconscious misdiagnosis of CRC based on their daily image analyses. MethodsBased on a state-of-the-art transfer-learned deep convolutional neural network in artificial intelligence (AI), we proposed a novel patch aggregation strategy for clinic CRC prediction/diagnosis using weakly labeled pathological whole slide image (WSI) patches. This approach was trained and validated using an unprecedented and enormously large number of 170,099 patches, >14,680 WSIs, from >9,631 subjects that covered diverse and representative clinical cases from multi-independent-sources across China, U.S., and Germany. ResultsOur innovative AI tool was consistently nearly perfectly agreed with (average Kappa-statistic 0.896) and even often better than most of the experienced expert pathologists when tested in diagnosing CRC WSIs from multi-centers. The average area under the receiver operating characteristics curve (AUC) of AI was greater than that of the pathologists (0.981 vs 0.970) and achieved the best performance among the application of other AI methods to CRC diagnosis. Our AI-generated heatmap highlights the image regions of cancer tissue/cells. ConclusionsThis first-ever generalizable AI system can handle large amounts of WSIs consistently and robustly without potential bias due to fatigue commonly experienced by clinical pathologists. Hence, it will drastically alleviate the heavy clinical burden of daily pathology diagnosis, and improve the treatment for CRC patients. This tool is generalizable to other cancer diagnosis based on image recognition.

pathology↗

bioRxiv2020

Nrf2 represses ER stress-related ferroptosis in renal carcinoma cells via HO-1

Selective-induction of regulated cell death is considered as a useful strategy for developing effective therapies against renal cell carcinoma (RCC). In this study, the role of nuclear factor-E2-related factor 2 (Nrf2) and its downstream genes in erastin-induced ferroptosis of renal carcinoma cells was investigated. Erastin induced endoplasmic reticulum (ER) stress and ferroptosis in renal carcinoma cells (786-0 and Caki-1), which was confirmed by morphological changes, increased lipid oxidation and ferroptosis inhibition. Data obtained from TCGA dataset showed that Nrf2 and heme oxygenase-1 (HO-1) were not only significantly differentially expressed between RCC and normal samples, but also associated with better survival in RCC patients. Moreover, the process of ferroptosis induced by erastin in renal carcinoma cells was accompanied by increased expression of Nrf2 and HO-1. Inhibition of Nrf2 or HO-1 expression enhanced lipid ROS accumulation and ferroptosis. Furthermore, pharmacological activation of HO-1 reversed ferroptosis-sensitizing effect of Nrf2 silencing, thereby clarifying the key role of Nrf2/HO-1 pathway in erastin-induced ferroptosis in renal carcinoma cells. In summary, we identified the role of Nrf2/HO-1 signaling pathway in the protection of renal carcinoma cells from ER stress-related ferroptosis, which indicated that targeting of Nrf2/HO-1 pathway as a viable treatment strategy for RCC. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=151 SRC="FIGDIR/small/977769v1_ufig1.gif" ALT="Figure 1"> View larger version (60K): org.highwire.dtl.DTLVardef@1068dacorg.highwire.dtl.DTLVardef@13f388eorg.highwire.dtl.DTLVardef@199fd87org.highwire.dtl.DTLVardef@d3e03a_HPS_FORMAT_FIGEXP M_FIG Graphical abstract (supplementary figure): Nrf2 protects against ER stress-related ferroptosis via HO-1 in renal cancer cells. Cystine/glutamate antiporter (SLC7A11) abbreviated as xCT. Mitochondria abbreviated as Mito. C_FIG

cancer biology↗

bioRxiv2020

The PK/PD Integration and Resistance of Ti lmi cosin Against Mycoplasma hyopneumoniae

Mycoplasma hyopneumoniae is the major pathogen causing enzootic pneumonia in pigs. M. hyopneumoniae infection can lead to considerable economic losses in the pig-breeding industry. Here, this study established a first-order absorption, one-compartment model to study the relationship between the pharmacokinetics/pharmacodynamics (PK/PD) index of tilmicosin against M. hyopneumoniae in vitro. We simulated the drug concentration of timicosin in the fluid lining the lung epithelia of pigs. The minimum inhibitory concentration (MIC) of tilmicosin against M. hyopneumoniae with an inoculum of 106 CFU/mL was 1.6 g/mL using the microdilution method. Static time-kill curves showed that, if the drug concentration >1 MIC, the antibacterial effect showed different degrees of inhibition. At 32 MIC, the amount of bacteria decreased by 3.16 log10 CFU/mL, thereby achieving a mycoplasmacidal effect. The M. hyopneumoniae count was reduced from 3.61 to 5.11 log10 CFU/mL upon incubation for 96 h in a dynamic model with a dose of 40-200 mg, thereby achieving mycoplasmacidal activity. The peak concentration by MIC (Cmax/MIC) and the area under the concentration-time curve over 96 h divided by the MIC (AUC0-96 h/MIC) were the best-fit PK/PD parameters for predicting the antibacterial activity of tilmicosin against M. hyopneumoniae (R2 = 0.99), suggesting that tilmicosin had concentration-dependent activity. The estimated value for Cmax/MIC and AUC0-96 h/MIC for 2log10 (CFU/mL) reduction and 3log10 (CFU/mL) reduction from baseline was 1.44 and 1.91, and 70.55 h and 96.72 h, respectively. Four M. hyopneumoniae strains (M1-M4) with reduced sensitivity to tilmicosin were isolated from the four dose groups. The susceptibility of these strains to tylosin, erythromycin and lincomycin was also reduced significantly. For sequencing analyses of 23S rRNA, an acquired A2058G transition in region V was found only in resistant M. hyopneumoniae strains (M3, M4). In conclusion, in an in vitro model, the effect of tilmicosin against M. hyopneumoniae was concentration-dependent and had a therapeutic effect. These results will help to design the optimal dosing regimen for tilmicosin in M. hyopneumoniae infection, and minimize the emergence of resistant bacteria.

pharmacology and toxicology↗

bioRxiv2020

RNA-seq and scRNA-seq reveal trajectory progression of the retinal ganglion cell lineage in wild-type and Atoh7-null retinas

Past studies concluded that Atoh7 is critical for the emergence of the retinal ganglion cell (RGC) lineage in the developing retina, whereas Pou4f2 and Isl1 function in RGC differentiation. Atoh7 is expressed in a subset of retinal progenitor cells (RPCs) and is considered a competence factor for the RGC fate, but the molecular properties of these RPCs have not been well characterized. In this study, we first used conventional RNA-seq to investigate transcriptomic changes in Atoh7-, Pou4f2-, and Isl1-null retinas at embryonic (E) day 14.5 and identified the differentially expressed genes (DEGs), which expanded our understanding of the scope of downstream events. We then performed single cell RNA-seq (scRNA-seq) on E13.5 and E17.5 wild-type and Atoh7-null retinal cells. Clustering analysis not only correctly identified known cell types at these developmental stages but also revealed a transitional cell state which was marked by Atoh7 and genes for other lineages in a highly overlapping fashion and shared by all early developmental trajectories. Further, analysis of the Atoh7-null retina revealed that, unlike previously believed, the RGC lineage still progressed considerably and a substantial amount of RGC-specific gene expression still occurred. Thus, Atoh7 likely collaborates with other factors to shepherd the transitional RPCs to the RGC lineage by competing with other lineage factors and activating RGC-specific genes. This study thus provides significant insights into the nature of RPC competence for different retinal cell fates and revises our current view on the emergence of the RGC lineage.

developmental biology↗

bioRxiv2020

A robust and interpretable, end-to-end deep learning model for cytometry data

Cytometry technologies are essential tools for immunology research, providing high-throughput measurements of the immune cells at the single-cell level. Traditional approaches in interpreting and using cytometry measurements include manual or automated gating to identify cell subsets from the cytometry data, providing highly intuitive results but may lead to significant information loss, in that additional details in measured or correlated cell signals might be missed. In this study, we propose and test a deep convolutional neural network for analyzing cytometry data in an end-to-end fashion, allowing a direct association between raw cytometry data and the clinical outcome of interest. Using nine large CyTOF studies from the open-access ImmPort database, we demonstrated that the deep convolutional neural network model can accurately diagnose the latent cytomegalovirus (CMV) in healthy individuals, even when using highly heterogeneous data from different studies. In addition, we developed a permutation-based method for interpreting the deep convolutional neural network model and identified a CD27-CD94+ CD8+ T cell population significantly associated with latent CMV infection. Finally, we provide a tutorial for creating, training and interpreting the tailored deep learning model for cytometry data using Keras and TensorFlow (github.com/hzc363/DeepLearningCyTOF).

bioinformatics↗

bioRxiv2020

Genome-wide mapping and prediction of plant architecture in a sorghum nested association mapping population

Modifying plant architecture is often necessary for yield improvement and climate adaptation, but we lack understanding of the genotype-phenotype map for plant morphology in sorghum. Here, we use a nested association mapping (NAM) population that captures global allelic diversity of sorghum to characterize the genetics of leaf erectness, leaf width (at two stages), and stem diameter. Recombinant inbred lines (n = 2200) were phenotyped in multiple environments (35,200 observations) and joint linkage mapping was performed with [~]93,000 markers. Fifty-four QTL of small to large effect were identified for trait BLUPs (9-16 per trait) each explaining 0.4-4% of variation across the NAM population. While some of these QTL colocalize with sorghum homologs of grass genes [e.g. involved in hormone synthesis (maize spi1), floral transition (SbCN8), and transcriptional regulation of development (rice Ideal plant architecture1)], most QTL did not colocalize with an a priori candidate gene (82%). Genomic prediction accuracy was generally high in five-fold cross-validation (0.65-0.83), and varied from low to high in leave-one-family-out cross-validation (0.04-0.61). The findings provide a foundation to identify the molecular basis of architecture variation in sorghum and establish genomic-enabled breeding for improved plant architecture. Core ideasO_LIUnderstanding the genetics of plant architecture could facilitate the development of crop ideotypes for yield and adaptation C_LIO_LIThe genetics of plant architecture traits was characterized in sorghum using multi-environment phenotyping in a global nested association mapping population C_LIO_LIFifty-five quantitative trait loci were identified; some colocalize with homologs of known developmental regulators but most do not C_LIO_LIGenomic prediction accuracy was consistently high in five-fold cross-validation, but accuracy varied considerably in leave-one-family-out predictions C_LI

genetics↗

bioRxiv2020

SIMPLEs: a single-cell RNA sequencing imputation strategy preserving gene modules and cell clusters variation

A main challenge in analyzing single-cell RNA sequencing (scRNASeq) data is to reduce technical variations yet retain cell heterogeneity. Due to low mRNAs content per cell and molecule losses during the experiment (called "dropout"), the gene expression matrix has substantial zero read counts. Existing imputation methods either treat each cell or each gene identically and independently, which oversimplifies the gene correlation and cell type structure. We propose a statistical model-based approach, called SIMPLEs, which iteratively identifies correlated gene modules and cell clusters and imputes dropouts customized for individual gene module and cell type. Simultaneously, it quantifies the uncertainty of imputation and cell clustering. Optionally, SIMPLEs can integrate bulk RNASeq data for estimating dropout rates. In simulations, SIMPLEs performed significantly better than prevailing scRNASeq imputation methods by various metrics. By applying SIMPLEs to several real data sets, we discovered gene modules that can further classify subtypes of cells. Our imputations successfully recovered the expression trends of marker genes in stem cell differentiation and can discover putative pathways regulating biological processes.

bioinformatics↗

bioRxiv2019

Benzaldehyde Synthases Are Encoded by Cinnamoyl-CoA Reductase Genes in Cucumber (Cucumis sativus L.)

Benzaldedyde, commonly detected in plant VOC (volatile organic compounds) profiling, is derived from phenylalanine. However, the last enzymatic step for benzaldedyde formation, designated as benzaldehyde synthase, remains elusive for long time. Here, we demonstrated that cinnamoyl-CoA reductases are responsible for benzaldedyde production in cucumber (Cucumis sativus L.). Comprehensive tissue specificity of VOC profiling revealed that benzaldehyde was specifically accumulated in root and flower of cucumber plants. VOC-gene correlation analysis suggested that several CCRs are candidate genes for benzaldehyde production: CsaCCR7 had a root-specific expression pattern while CsaCCR9 and CsaCCR18 showed a flower-specific expression pattern. Enzymatic assay demonstrated that CsaCCR7, CsaCCR9 and CsaCCR18 convert benzoyl-CoA to benzaldehyde. Subcellular localization experiments revealed that CsaCCR7 and CsaCCR18 are localized in cytosol, while CsaCCR9 was localized in peroxisome. In contrast to the long-standing view that CCR enzymes are involved in lignin biosynthesis in plants, it is the first time here to add a new biochemical role of CCR as benzaldehyde synthase in plants. HighlightsBenzaldehyde is mainly produced in flower and root of cucumber plants. 14 genes encoding CCR enzyme from cucumber are comprehensively analyzed. Three CsaCCRs, function as benzaldehyde synthases, utilize benzoyl-CoA as substrate to produce benzaldehyde in vitro.

biochemistry↗

bioRxiv2019

Delineating the Organization of Projection Neuron Subsets with Multi-fluorescent Rabies Virus Tracing Tool

The elegant functions of the brain are facilitated by sophisticated connections between neurons, the architecture of which is frequently characterized by one nucleus connecting to multiple targets via projection neurons. Delineating the sub-nucleus fine architecture of projection neurons in a certain nucleus could greatly facilitate its circuit, computational, and functional resolution. Here, we developed multi-fluorescent rabies virus to delineate the fine organization of corticothalamic projection neuron subsets in the primary visual cortex (V1). By simultaneously labeling multiple distinct subsets of corticothalamic projection neurons in V1 from their target nuclei in thalamus (dLGN, LP, LD), we observed that V1-dLGN corticothalamic neurons were densely concentrated in layer VI, except for several sparsely scattered neurons in layer V, while V1-LP and V1-LD corticothalamic neurons were localized to both layers V and VI. Meanwhile, we observed a fraction of V1 corticothalamic neurons targeting multiple thalamic nuclei, which was further confirmed by fMOST whole-brain imaging. We further conceptually proposed an upgraded sub-nucleus tracing system with higher throughput (21 subsets) for more complex architectural tracing. The multi-fluorescent RV tracing tool can be extensively applied to resolve architecture of projection neuron subsets, with a strong potential to delineate the computational and functional organization of these nuclei.

neuroscience↗

bioRxiv2019

Pan-cancer analysis of clonality and the timing of systemic spread in paired primary tumors and metastases

Metastasis is the primary cause of cancer-related deaths, but the natural history, clonal evolution and impact of treatment are poorly understood. We analyzed exome sequencing data from 457 paired primary tumor and metastatic samples from 136 breast, colorectal and lung cancer patients, including untreated (n=99) and treated (n=100) metastatic tumors. Treated metastases often harbored private driver mutations whereas untreated metastases did not, suggesting that treatment promotes clonal evolution. Polyclonal seeding was common in untreated lymph node metastases (n=17/29, 59%) and distant metastases (n=20/70, 29%), but less frequent in treated distant metastases (n=9/94, 10%). The low number of metastasis-private clonal mutations is consistent with early metastatic seeding, which we estimated commonly occurred 2-4 years prior to diagnosis across these cancers. Further, these data suggest that the natural course of metastasis is selectively relaxed relative to early tumor development and that metastasis-private mutations are not drivers of cancer spread but instead associated with drug resistance.

cancer biology↗

bioRxiv2019

Integrating multi-omics data with deep learning for predicting cancer prognosis

BackgroundGenomic information is nowadays widely used for precise cancer treatments. Since the individual type of omics data only represents a single view that suffers from data noise and bias, multiple types of omics data are required for accurate cancer prognosis prediction. However, it is challenging to effectively integrate multi-omics data due to the large number of redundant variables but relatively small sample size. With the recent progress in deep learning techniques, Autoencoder was used to integrate multi-omics data for extracting representative features. Nevertheless, the generated model is fragile from data noises. Additionally, previous studies usually focused on individual cancer types without making comprehensive tests on pan-cancer. Here, we employed the denoising Autoencoder to get a robust representation of the multi-omics data, and then used the learned representative features to estimate patients risks. ResultsBy applying to 15 cancers from The Cancer Genome Atlas (TCGA), our method was shown to improve the C-index values over previous methods by 6.5% on average. Considering the difficulty to obtain multi-omics data in practice, we further used only mRNA data to fit the estimated risks by training XGboost models, and found the models could achieve an average C-index value of 0.627. As a case study, the breast cancer prognosis prediction model was independently tested on three datasets from the Gene Expression Omnibus (GEO), and shown able to significantly separate high-risk patients from low-risk ones (C-index>0.6, p-values<0.05). Based on the risk subgroups divided by our method, we identified nine prognostic markers highly associated with breast cancer, among which seven genes have been proved by literature review. ConclusionOur comprehensive tests indicated that we have constructed an accurate and robust framework to integrate multi-omics data for cancer prognosis prediction. Moreover, it is an effective way to discover cancer prognosis-related genes.

bioinformatics↗