Biology subjects

Serafim Batzoglou

Publications and source records attributed to Serafim Batzoglou.

4 recordsLinked to original sources

bioRxiv2016

Genome-wide reconstruction of complex structural variants using read clouds

Recently developed methods that utilize partitioning of long genomic DNA fragments, and barcoding of shorter fragments derived from them, have succeeded in retaining long-range information in short sequencing reads. These so-called read cloud approaches represent a powerful, accurate, and cost-effective alternative to single-molecule long-read sequencing. We developed software, GROC-SVs, that takes advantage of read clouds for structural variant detection and assembly. We apply the method to two 10x Genomics data sets, one chromothriptic sarcoma with several spatially separated samples, and one breast cancer cell line, all Illumina-sequenced to high coverage. Comparison to short-fragment data from the same samples, and validation by mate-pair data from a subset of the sarcoma samples, demonstrate substantial improvement in specificity of breakpoint detection compared to short-fragment sequencing, at comparable sensitivity, and vice versa. The embedded longrange information also facilitates sequence assembly of a large fraction of the breakpoints; importantly, consecutive breakpoints that are closer than the average length of the input DNA molecules can be assembled together and their order and arrangement reconstructed, with some events exhibiting remarkable complexity. These features facilitated an analysis of the structural evolution of the sarcoma. In the chromothripsis, rearrangements occurred before copy number amplifications, and using the phylogenetic tree built from point mutation data we show that single nucleotide variants and structural variants are not correlated. We predict significant future advances in structural variant science using 10x data analyzed with GROC-SVs and other read cloud-specific methods.

Genomics↗

bioRxiv2016

Visualization and analysis of single-cell RNA-seq data by kernel-based similarity learning

Single-cell RNA-seq technologies enable high throughput gene expression measurement of individual cells, and allow the discovery of heterogeneity within cell populations. Measurement of cell-to-cell gene expression similarity is critical to identification, visualization and analysis of cell populations. However, single-cell data introduce challenges to conventional measures of gene expression similarity because of the high level of noise, outliers and dropouts. Here, we propose a novel similarity-learning framework, SIMLR (single-cell interpretation via multi-kernel learning), which learns an appropriate distance metric from the data for dimension reduction, clustering and visualization applications. Benchmarking against state-of-the-art methods for these applications, we used SIMLR to re-analyse seven representative single-cell data sets, including high-throughput droplet-based data sets with tens of thousands of cells. We show that SIMLR greatly improves clustering sensitivity and accuracy, as well as the visualization and interpretability of the data.

Bioinformatics↗

bioRxiv2016

Privacy-Preserving Read Mapping Using Locality Sensitive Hashing and Secure Kmer Voting

The recent explosion in the amount of available genome sequencing data imposes high computational demands on the tools designed to analyze it. Low-cost cloud computing has the potential to alleviate this burden. However, moving personal genome data analysis to the cloud raises serious privacy concerns. Read alignment is a critical and computationally intensive first step of most genomic data analysis pipelines. While significant effort has been dedicated to optimize the sensitivity and runtime efficiency of this step, few approaches have addressed outsourcing this computation securely to an untrusted party. The few secure solutions that have been proposed either do not scale to whole genome sequencing datasets or are not competitive with the state of the art in read mapping. In this paper, we present BALAUR, a privacy-preserving read mapping algorithm based on locality sensitive hashing and secure kmer voting. BALAUR securely outsources a significant portion of the computation to the public cloud by formulating the alignment task as a voting scheme between encrypted read and reference kmers. Our approach can easily handle typical genome-scale datasets and is highly competitive with non-cryptographic state-of-the-art read aligners in both accuracy and runtime performance on simulated and real read data. Moreover, our approach is significantly faster than state-of-the-art read aligners in long read mapping.

Bioinformatics↗

bioRxiv2014

Reveel: large-scale population genotyping using low-coverage sequencing data

Population low-coverage whole-genome sequencing is rapidly emerging as a prominent approach for discovering genomic variation and genotyping a cohort. This approach combines substantially lower cost than full-coverage sequencing with whole-genome discovery of low-allele-frequency variants, to an extent that is not possible with array genotyping or exome sequencing. However, a challenging computational problem arises when attempting to discover variants and genotype the entire cohort. Variant discovery and genotyping are relatively straightforward on a single individual that has been sequenced at high coverage, because the inference decomposes into the independent genotyping of each genomic position for which a sufficient number of confidently mapped reads are available. However, in cases where low-coverage population data are given, the joint inference requires leveraging the complex linkage disequilibrium patterns in the cohort to compensate for sparse and missing data in each individual. The potentially massive computation time for such inference, as well as the missing data that confound low-frequency allele discovery, need to be overcome for this approach to become practical. Here, we present Reveel, a novel method for single nucleotide variant calling and genotyping of large cohorts that have been sequenced at low coverage. Reveel introduces a novel technique for leveraging linkage disequilibrium that deviates from previous Markov-based models. We evaluate Reveels performance through extensive simulations as well as real data from the 1000 Genomes Project, and show that it achieves higher accuracy in low-frequency allele discovery and substantially lower computation cost than previous state-of-the-art methods.

Bioinformatics↗