Biology subjects

Zhi, D.

Publications and source records attributed to Zhi, D..

43 records · Page 3Linked to original sources

bioRxiv2019

FUNCTIONAL NETWORK CONNECTIVITY (FNC)-BASED GENERATIVE ADVERSARIAL NETWORK (GAN) AND ITS APPLICATIONS IN MENTAL DISORDERS DIAGNOSIS

Functional network connectivity (FNC) obtained from resting-state functional magnetic resonance imaging (fMRI) data have been commonly used to study mental disorders in neuroimaging applications. Likewise, generative adversarial networks (GANs) have performed well in multiple classification benchmark tasks. However, the application of GANs to fMRI is relatively rare. In this work, we proposed an FNC-based GAN for classifying brain disorders from healthy controls (HCs), in which FNC matrices were calculated by correlation of time courses derived from non-artefactual fMRI independent components (ICs). The proposed GAN model consisted of one discriminator (real FNCs) and one generator (fake FNCs), each has four fully-connected layers, and feature matching was implemented between each other to improve classification performance. An average accuracy of 70.1% with 10-fold cross-validation was achieved for classifying 269 major depressive disorder (MDD) patients from 286 HCs, at least 5.9% higher compared to other 6 popular classification approaches (54.5-64.2%). In another application to discriminating between 558 schizophrenia patients and 542 HCs from 7 sites, the proposed GAN model achieved 80.7% accuracy in leave-one-site-out prediction, outperforming support vector machine (SVM) and deep neural net (DNN) by 3-6%. To the best of our knowledge, this is the first attempt to apply GAN model based on fMRI data for mental disorder classification. Such a framework promises wide utility and great potential in neuroimaging biomarker identification.

bioengineering↗

bioRxiv2019

HaploHide: A Data Hiding Framework for Privacy Enhanced Sharing of Personal Genetic Data

Personal genetic data is becoming a digital commodity as millions of individuals have direct access to and control of their genetic information. This information must be protected as it can be used for reidentification and potential discrimination of individuals and relatives. While there is a great incentive to share and use genetic information, there are limited number of practical approaches for protecting it when individuals would like to make use of their genomes in clinical and recreational settings. To enable privacy-enhanced usage of genomic data by individuals, we propose a crowd-blending-based framework where portions of the individuals haplotype is \"hidden\" within a large sample of other haplotypes. The hiding framework is motivated by the existence of large-scale population panels that we utilize for generation of the crowd of haplotypes in which the individuals haplotype is hidden. We demonstrate the usage of hiding in two different scenarios: Sharing of variant alleles on genes and sharing of GWAS variant alleles. We evaluate hiding framework by testing reidentification of hidden individuals using numerous measures of individual reidentification. In these settings, we discuss how effective hiding can be accomplished when the adversary does not have access to auxiliary identifying information. Compared to the existing approaches for protecting privacy, which require substantial changes in the computational infrastructure, e.g., homomorphic encryption, hiding-based framework does not incur any changes to the infrastructure. However, the processing must be performed for every sample in the crowd and therefore data processing cost will increase as the crowd size increases.

bioinformatics↗

bioRxiv2019

Parallel Group ICA + ICA: Joint Estimation of Linked Functional Network Variability and Structural Covariation with Application to Schizophrenia

There is growing evidence that rather than using a single brain imaging modality to study its association with physiological or symptomatic features, the field is paying more attention to fusion of multimodal information. However, most current multimodal fusion approaches that incorporate functional magnetic resonance imaging (fMRI) are restricted to second-level 3D features, rather than the original 4D fMRI data. This trade-off is that the valuable temporal information is not utilized during the fusion step. Here we are motivated to propose a novel approach called \"parallel group ICA+ICA\" that incorporates temporal fMRI information from GICA into a parallel ICA framework, aiming to enable direct fusion of first-level fMRI features with other modalities (e.g. structural MRI), which thus can detect linked functional network variability and structural covariations. Simulation results show that the proposed method yields accurate inter-modality linkage detection regardless of whether it is strong or weak. When applied to real data, we identified one pair of significantly associated fMRI-sMRI components that show group difference between schizophrenia and controls in both modalities. Finally, multiple cognitive domain scores can be predicted by the features identified in the linked component pair by our proposed method. We also show these multimodal brain features can predict multiple cognitive scores in an independent cohort. Overall, results demonstrate the ability of parallel GICA+ICA to estimate joint information from 4D and 3D data without discarding much of the available information up front, and the potential for using this approach to identify imaging biomarkers to study brain disorders.

bioinformatics↗

bioRxiv2018

Leveraging collective regulatory effects of long-range DNA methylations to predict gene expressions and estimate their effects on phenotypes in cancer

DNA methylation of various genomic regions plays an important role in regulating gene expression in diverse biological contexts. However, most genome-wide studies have focused on the effect of 1) methylation in cis, not in trans and 2) a single CpG, not the collective effects of multiple CpGs, on gene expression. In this study, we developed a statistical machine learning model, geneEXPLORER (gene expression prediction by long-range epigenetic regulation), that quantifies the collective effects of both cis- and trans- methylations on gene expression. By applying geneEXPLORER to The Cancer Genome Atlas (TCGA) breast and lung cancer data, we found that most genes are affected by methylations of as much as 10Mb from promoter regions or more, and the long-range methylation explains 50% of the variation in gene expression on average, far greater than cis-methylation. The highly predictive genes are related to breast cancer, especially oncogenes and suppressor genes. Further, the predicted gene expressions could predict clinical phenotypes such as breast tumor status and estrogen receptor status (AUC=0.999, 0.94 respectively) as accurately as the measured gene expression levels. These results suggest that geneEXPLORER provides a means for accurate imputation of gene expression, which can be further used to predict clinical phenotypes.

bioinformatics↗

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↗

bioRxiv2017

Ultra-fast Identity by Descent Detection in Biobank-Scale Cohorts using Positional Burrows-Wheeler Transform

With the availability of genotyping data of very large samples, there is an increasing need for tools that can efficiently identify genetic relationships among all individuals in the sample. One fundamental measure of genetic relationship of a pair of individuals is identity by descent (IBD), chromosomal segments that are shared among two individuals due to common ancestry. However, the efficient identification of IBD segments among a large number of genotyped individuals is a challenging computational problem. Most existing methods are not feasible for even thousands of individuals because they are based on pairwise comparisons of all individuals and thus scale up quadratically with sample size. Some methods, such as GERMLINE, use fast dictionary lookup of short seed sequence matches to achieve a near-linear time efficiency. However, the number of short seed matches often scales up super-linearly in real population data.\n\nIn this paper we describe a new approach for IBD detection. We take advantage of an efficient population genotype index, Positional BWT (PBWT), by Richard Durbin. PBWT achieves linear time query of perfectly identical subsequences among all samples. However, the original PBWT is not tolerant to genotyping errors which often interrupt long IBD segments into short fragments. We introduce a randomized strategy by running PBWTs over random projections of the original sequences. To boost the detection power we run PBWT multiple times and merge the identified IBD segments through interval tree algorithms. Given a target IBD segment length, RaPID adjust parameters to optimize detection power and accuracy.\n\nSimulation results proved that our tool (RaPID) achieves almost linear scaling up to sample size and is orders of magnitude faster than GERMLINE. At the same time, RaPID maintains a detection power and accuracy comparable to existing mainstream algorithms, GERMLINE and IBDseq. Running multiple times with various target detection lengths over the 1000 Genomes Project data, RaPID can detect population events at different time scales. With our tool, it is feasible to identify IBDs among hundreds of thousands to millions of individuals, a sample size that will become reality in a few years.

genomics↗

bioRxiv2016

Hardy Weinberg Exact Test In Large Scale Variant Calling Quality Control

Hardy Weinberg Equilibrium (HWE) test is widely used as a quality control measure to detect sequencing artifacts like mismapping, allelic dropout and biases. However, in the high throughput sequencing era, where the sample size is beyond a thousand scale, the utility of HWE test in reducing the false positive rate remains unclear. In this paper, we demonstrate that HWE test has limited power in identifying sequencing artifacts when the variant allele frequency is lower than 1% in a variant call set produced from more than five thousand whole genome sequenced samples from two homogeneous populations. We develop a novel strategy of implementing HWE filtering in which we incorporate site frequency spectrum information and determine the p-value cutoff which optimizes the tradeoff between sensitivity and specificity. The novel strategy is shown to outperform the exact test of HWE with an empirical constant p-value cutoff regardless of the sequencing sample size. We also present best practice recommendations for identifying possible sources of false positives from large sequencing datasets based on an analysis of intrinsic biases in the variant calling process. Our novel strategy of determining the HWE test p-value cutoff and applying the test to the common variants provides a practical approach for the variant level quality controls in the upcoming sequencing projects with tens to hundreds of thousand of samples.

bioinformatics↗