Biology subjects

Chang, P.-C.

Publications and source records attributed to Chang, P.-C..

29 records · Page 2Linked to original sources

bioRxiv2021

A Complete Pedigree-Based Graph Workflow for Rare Candidate Variant Analysis

Methods that use a linear genome reference for genome sequencing data analysis are reference biased. In the field of clinical genetics for rare diseases, a resulting reduction in genotyping accuracy in some regions has likely prevented the resolution of some cases. Pangenome graphs embed population variation into a reference structure. While pangenome graphs have helped to reduce reference mapping bias, further performance improvements are possible. We introduce VG-Pedigree, a pedigree-aware workflow based on the pangenome-mapping tool of Giraffe (Siren et al. 2021) and the variant-calling tool DeepTrio (Kolesnikov et al. 2021) using a specially-trained model for Giraffe-based alignments. We demonstrate mapping and variant calling improvements in both single-nucleotide variants (SNVs) and insertion and deletion (INDEL) variants over those produced by alignments created using BWA-MEM to a linear-reference and Giraffe mapping to a pangenome graph containing data from the 1000 Genomes Project. We have also adapted and upgraded the deleterious-variant (DV) detecting methods and programs of Gu et al. into a streamlined workflow (Gu et al. 2019). We used these workflows in combination to detect small lists of candidate DVs among 15 family quartets and quintets of the Undiagnosed Diseases Program (UDP). All candidate DVs that were previously diagnosed using the mendelian models covered by the previously published Gu et al. methods were recapitulated by these workflows. The results of these experiments indicate a slightly greater absolute count of DVs are detected in the proband population than in their matched unaffected siblings.

bioinformatics↗

bioRxiv2021

DeepConsensus: Gap-Aware Sequence Transformers for Sequence Correction

Pacific BioScience (PacBio) circular consensus sequencing (CCS) generates long (10-25 kb), accurate "HiFi" reads by combining serial observations of a DNA molecule into a consensus sequence. The standard approach to consensus generation uses a hidden Markov model (pbccs). Here, we introduce DeepConsensus, which uses a unique alignment-based loss to train a gap-aware transformer-encoder (GATE) for sequence correction. Compared to pbccs, DeepConsensus reduces read errors in the same dataset by 42%. This increases the yield of PacBio HiFi reads at Q20 by 9%, at Q30 by 27%, and at Q40 by 90%. With two SMRT Cells of HG003, reads from DeepConsensus improve hifiasm assembly contiguity (NG50 4.9Mb to 17.2Mb), increase gene completeness (94% to 97%), reduce false gene duplication rate (1.1% to 0.5%), improve assembly base accuracy (Q43 to Q45), and also reduce variant calling errors by 24%.

bioinformatics↗

bioRxiv2021

DeepTrio: Variant Calling in Families Using Deep Learning

Every human inherits one copy of the genome from their mother and another from their father. Parental inheritance helps us understand the transmission of traits and genetic diseases, which often involve de novo variants and rare recessive alleles. Here we present DeepTrio, which learns to analyze child-mother-father trios from the joint sequence information, without explicit encoding of inheritance priors. DeepTrio learns how to weigh sequencing error, mapping error, and de novo rates and genome context directly from the sequence data. DeepTrio has higher accuracy on both Illumina and PacBio HiFi data when compared to DeepVariant. Improvements are especially pronounced at lower coverages (with 20x DeepTrio roughly equivalent to 30x DeepVariant). DeepTrio includes pre-trained models for Illumina WGS, Illumina exome, and PacBio HiFi.

bioinformatics↗

bioRxiv2021

Haplotype-aware variant calling enables high accuracy in nanopore long-reads using deep neural networks

Long-read sequencing has the potential to transform variant detection by reaching currently difficult-to-map regions and routinely linking together adjacent variations to enable read based phasing. Third-generation nanopore sequence data has demonstrated a long read length, but current interpretation methods for its novel pore-based signal have unique error profiles, making accurate analysis challenging. Here, we introduce a haplotype-aware variant calling pipeline PEPPER-Margin-DeepVariant that produces state-of-the-art variant calling results with nanopore data. We show that our nanopore-based method outperforms the short-read-based single nucleotide variant identification method at the whole genome-scale and produces high-quality single nucleotide variants in segmental duplications and low-mappability regions where short-read based genotyping fails. We show that our pipeline can provide highly-contiguous phase blocks across the genome with nanopore reads, contiguously spanning between 85% to 92% of annotated genes across six samples. We also extend PEPPER-Margin-DeepVariant to PacBio HiFi data, providing an efficient solution with superior performance than the current WhatsHap-DeepVariant standard. Finally, we demonstrate de novo assembly polishing methods that use nanopore and PacBio HiFi reads to produce diploid assemblies with high accuracy (Q35+ nanopore-polished and Q40+ PacBio-HiFi-polished).

bioinformatics↗

bioRxiv2021

Improving variant calling using population data and deep learning

Large-scale population variant data is often used to filter and aid interpretation of variant calls in a single sample. These approaches do not incorporate population information directly into the process of variant calling, and are often limited to filtering which trades recall for precision. In this study, we develop population-aware DeepVariant models with a new channel encoding allele frequencies from the 1000 Genomes Project. This model reduces variant calling errors, improving both precision and recall in single samples, and reduces rare homozygous and pathogenic clinvar calls cohort-wide. We assess the use of population-specific or diverse reference panels, finding the greatest accuracy with diverse panels, suggesting that large, diverse panels are preferable to individual populations, even when the population matches sample ancestry. Finally, we show that this benefit generalizes to samples with different ancestry from the training data even when the ancestry is also excluded from the reference panel.

bioinformatics↗

bioRxiv2020

An Extensive Sequence Dataset of Gold-Standard Samples for Benchmarking and Development

Accurate standards and extensive development datasets are the foundation of technical progress. To facilitate benchmarking and development, we sequence 9 samples, covering the Genome in a Bottle truth sets on multiple instruments (NovaSeq, HiSeqX, HiSeq4000, PacBio Sequel II System) and sample preparations (PCR-Free, PCR-Positive) for both whole genome and multiple exome kits. We benchmark pipelines, quantifying strengths and limitations for sequencing and analysis methods. We identify variability within and between instruments, preparation methods, and analytical pipelines, across various sequencing depths. We discuss the relevance of this variability to downstream analyses, and strategies to reduce variability.

genomics↗

bioRxiv2020

Genotyping common, large structural variations in 5,202 genomes using pangenomes, the Giraffe mapper, and the vg toolkit

We introduce Giraffe, a pangenome short read mapper that can efficiently map to a collection of haplotypes threaded through a sequence graph. Giraffe, part of the variation graph toolkit (vg)1, maps reads to thousands of human genomes at around the same speed BWA-MEM2 maps reads to a single reference genome, while maintaining comparable accuracy to VG-MAP, vgs original mapper. We have developed efficient genotyping pipelines using Giraffe. We demonstrate improvements in genotyping for single-nucleotide variants (SNVs), small insertions and deletions (indels) and structural variations (SVs) genome-wide. We use Giraffe to genotype about 167 thousand structural variants ascertained from long read studies in 5,202 human genomes sequenced with short reads, including the complete 1000 Genomes Project dataset, at an average cost of $1.50 per sample. We determine the frequency of these variations in diverse human populations, characterize their complex allelic variations and identify thousands of expression quantitative trait loci (eQTLs) driven by these variations.

bioinformatics↗

bioRxiv2020

precisionFDA Truth Challenge V2: Calling variants from short- and long-reads in difficult-to-map regions

The precisionFDA Truth Challenge V2 aimed to assess the state-of-the-art of variant calling in difficult-to-map regions and the Major Histocompatibility Complex (MHC). Starting with FASTQ files, 20 challenge participants applied their variant calling pipelines and submitted 64 variant callsets for one or more sequencing technologies (~35X Illumina, ~35X PacBio HiFi, and ~50X Oxford Nanopore Technologies). Submissions were evaluated following best practices for benchmarking small variants with the new GIAB benchmark sets and genome stratifications. Challenge submissions included a number of innovative methods for all three technologies, with graph-based and machine-learning methods scoring best for short-read and long-read datasets, respectively. New methods out-performed the 2016 Truth Challenge winners, and new machine-learning approaches combining multiple sequencing technologies performed particularly well. Recent developments in sequencing and variant calling have enabled benchmarking variants in challenging genomic regions, paving the way for the identification of previously unknown clinically relevant variants.

bioinformatics↗

bioRxiv2020

Accurate, scalable cohort variant calls using DeepVariant and GLnexus

Population-scale sequenced cohorts are foundational resources for genetic analyses, but processing raw reads into analysis-ready variants remains challenging. Here we introduce an open-source cohort variant-calling method using the highly-accurate caller DeepVariant and scalable merging tool GLnexus. We optimized callset quality based on benchmark samples and Mendelian consistency across many sample sizes and sequencing specifications, resulting in substantial quality improvements and cost savings over existing best practices. We further evaluated our pipeline in the 1000 Genomes Project (1KGP) samples, showing superior quality metrics and imputation performance. We publicly release the 1KGP callset to foster development of broad studies of genetic variation.

genomics↗

bioRxiv2019

Highly-accurate long-read sequencing improves variant detection and assembly of a human genome

The major DNA sequencing technologies in use today produce either highly-accurate short reads or noisy long reads. We developed a protocol based on single-molecule, circular consensus sequencing (CCS) to generate highly-accurate (99.8%) long reads averaging 13.5 kb and applied it to sequence the well-characterized human HG002/NA24385. We optimized existing tools to comprehensively detect variants, achieving precision and recall above 99.91% for SNVs, 95.98% for indels, and 95.99% for structural variants. We estimate that 2,434 discordances are correctable mistakes in the high-quality Genome in a Bottle benchmark. Nearly all (99.64%) variants are phased into haplotypes, which further improves variant detection. De novo assembly produces a highly contiguous and accurate genome with contig N50 above 15 Mb and concordance of 99.998%. CCS reads match short reads for small variant detection, while enabling structural variant detection and de novo assembly at similar contiguity and markedly higher concordance than noisy long reads.

genomics↗

bioRxiv2018

A deep learning approach to pattern recognition for short DNA sequences

MotivationInferring properties of biological sequences--such as determining the species-of-origin of a DNA sequence or the function of an amino-acid sequence--is a core task in many bioinformatics applications. These tasks are often solved using string-matching to map query sequences to labeled database sequences or via Hidden Markov Model-like pattern matching. In the current work we describe and assess an deep learning approach which trains a deep neural network (DNN) to predict database-derived labels directly from query sequences. ResultsWe demonstrate this DNN performs at state-of-the-art or above levels on a difficult, practically important problem: predicting species-of-origin from short reads of 16S ribosomal DNA. When trained on 16S sequences of over 13,000 distinct species, our DNN achieves read-level species classification accuracy within 2.0% of perfect memorization of training data, and produces more accurate genus-level assignments for reads from held-out species than k-mer, alignment, and taxonomic binning baselines. Moreover, our models exhibit greater robustness than these existing approaches to increasing noise in the query sequences. Finally, we show that these DNNs perform well on experimental 16S mock community dataset. Overall, our results constitute a first step towards our long-term goal of developing a general-purpose deep learning approach to predicting meaningful labels from short biological sequences. AvailabilityTensorFlow training code is available through GitHub (https://github.com/tensorflow/models/tree/master/research). Data in TensorFlow TFRecord format is available on Google Cloud Storage (gs://brain-genomics-public/research/seq2species/). Contactseq2species-interest@google.com Supplementary informationSupplementary data are available in a separate document.

bioinformatics↗