Biology subjects

Dai, Y.

Publications and source records attributed to Dai, Y..

4 recordsLinked to original sources

bioRxiv2018

Gene2Vec: Distributed Representation of Genes Based on Co-Expression

Existing functional description of genes are categorical, discrete, and mostly through manual process. In this work, we explore the idea of gene embedding, distributed representation of genes, in the spirit of word embedding. From a pure data-driven fashion, we trained a 300 dimension vector representation of all human genes, using gene co-expression patterns in 984 data sets from the GEO databases. These vectors capture functional relatedness of genes in terms of recovering known pathways - the average inner product (similarity) of genes within a pathway is 1.68X greater than that of random genes. Using t-SNE, we produced a gene co-expression map that shows local concentrations of tissue specific genes. We also illustrated the usefulness of the embedded gene vectors, laden with rich information on gene co-expression patterns, in tasks such as gene-gene interaction prediction. Overall, we believe that this distributed representation of genes may be useful for more bioinformatics applications.

bioinformatics↗

bioRxiv2018

Single-molecule optical mapping enables accurate molecular diagnosis of facioscapulohumeral muscular dystrophy (FSHD)

Facioscapulohumeral Muscular Dystrophy (FSHD) is a common adult muscular dystrophy in which the muscles of the face, shoulder blades and upper arms are among the most affected. FSHD is the only disease in which \"junk\" DNA is reactivated to cause disease, and the only known repeat array-related disease where fewer repeats cause disease. More than 95% of FSHD cases are associated with copy number loss of a 3.3kb tandem repeat (D4Z4 repeat) at the subtelomeric chromosomal region 4q35, of which the pathogenic allele contains less than 10 repeats and has a specific genomic configuration called 4qA. Currently, genetic diagnosis of FSHD requires pulsed-field gel electrophoresis followed by Southern blot, which is labor-intensive, semi-quantitative and requires long turnaround time. Here, we developed a novel approach for genetic diagnosis of FSHD, by leveraging Bionano Saphyr single-molecule optical mapping platform. Using a bioinformatics pipeline developed for this assay, we found that the method gives direct quantitative measurement of repeat numbers, can differentiate 4q35 and the highly paralogous 10q26 regions, can determine the 4qA/4qB allelic configuration, and can quantitate levels of post-zygotic mosaicism. We evaluated this approach on 5 patients (including two with post-zygotic mosaicism) and 2 patients (including one with post-zygotic mosaicism) from two separate cohorts, and had complete concordance with Southern blots, but with improved quantification of repeat numbers resolved between haplotypes. We concluded that single-molecule optical mapping is a viable approach for molecular diagnosis of FSHD and may be applied in clinical diagnostic settings once more validations are performed.

genomics↗

bioRxiv2018

PopPhy-CNN: A Phylogenetic Tree Embedded Architecture for Convolution Neural Networks for Metagenomic Data

MotivationAccurate prediction of the host phenotype from a metgenomic sample and identification of the associated bacterial markers are important in metagenomic studies. We introduce PopPhy-CNN, a novel convolutional neural networks (CNN) learning architecture that effectively exploits phylogentic structure in microbial taxa. PopPhy-CNN provides an input format of 2D matrix created by embedding the phylogenetic tree that is populated with the relative abundance of microbial taxa in a metagenomic sample. This conversion empowers CNNs to explore the spatial relationship of the taxonomic annotations on the tree and their quantitative characteristics in metagenomic data.\n\nResultsPopPhy-CNN is evaluated using three metagenomic datasets of moderate size. We show the superior performance of PopPhy-CNN compared to random forest, support vector machines, LASSO and a baseline 1D-CNN model constructed with relative abundance microbial feature vectors. In addition, we design a novel scheme of feature extraction from the learned CNN models and demonstrate the improved performance when the extracted features are used to train support vector machines.\n\nConclusionPopPhy-CNN is a novel deep learning framework for the prediction of host phenotype from metagenomic samples. PopPhy-CNN can efficiently train models and does not require excessive amount of data. PopPhy-CNN facilities not only retrieval of informative microbial taxa from the trained CNN models but also visualization of the taxa on the phynogenetic tree.\n\nContactyagndai@uic.edu\n\nAvailabilitySource code is publicly available at https://github.com/derekreiman/PopPhy-CNN\n\nSupplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗

bioRxiv2017

MeDEStrand: an improved method to infer genome-wide absolute methylation level from DNA enrichment experiment

BackgroundDNA methylation of dinucleotide CpG is an essential epigenetic modification that plays a key role in transcription. Bisulfite conversion method is a \"gold standard\" for DNA methylation profiling that provides single nucleotide resolution. However, whole-genome bisulfite conversion is very expensive. Alternatively, DNA enrichment-based methods offer high coverage of methylated CpG dinucleotides with the lowest cost per CpG covered genome-wide and have been used widely. They measure the DNA enrichment of methyl-CpG binding, therefore do not directly provide absolute methylation levels. Further, the enrichment is influenced by confounding factors besides the methylation status, e.g., CpG density. Computational models that can accurately derive the absolute methylation levels from the enrichment data are necessary.\n\nResultsWe present MeDEStrand, a method uses sigmoid function to estimate and correct the CpG bias from the numbers of reads that fell within bins that divide the genome. In addition, unlike the previous methods, which estimate CpG bias based on reads mapped at the same genomic loci, MeDEStrand processes the reads for the positive and negative DNA strands separately. We compare the performance of MeDEStrand with three other state-of-the-art methods MEDIPS, BayMeth and QSEA on four independent datasets generated using immortalized cell lines (GM12878 and K562) and human patient primary cells (foreskin fibroblast and mammary epithelial). Based on the comparison between the inferred absolute methylation levels from MeDIP-seq and the corresponding RRBS data, MeDEStrand shows the best performance at high resolution of 25, 50 and 100 base pairs.\n\nConclusions MeDEStrand benefits from the estimation of CpG bias with a sigmoid function and the procedure to process reads mapped to the positive and negative DNA strands separately. MeDEStrand is a tool to infer whole-genome absolute DNA methylation level at the cost of enrichment-based methods with adequate accuracy and resolution. R package MeDEStrand and its tutorial is freely available for download at https://github.com/jxu1234/MeDEStrand.git

bioinformatics↗