11 research outputs found
Identification of structural variations from whole genome sequencing of cancer patients
No full textCancer is largely driven by accumulation of somatic mutations that can be subdivided into small mutations (single nucleotide variations (SNVs), small insertions and deletions) and large structural variations (SVs). While SNVs affect single nucleotide, SVs can affect large stretches of DNA. Reliable identification of all mutations is key to understanding genetic diseases like cancer.
SVs can be identified by whole genome sequencing with conventional Illumina short-read sequencing (cWGS) being the most widely used approach. However, reliable prediction of SVs with short-reads (50-150bp) from fragmented DNA (~0.5kb) is challenging due to ambiguous mapping reads at repetitive regions and typically only few short reads span rearranged SV breakpoints with limited sequence overlap (due to read length). The 10X Genomics linked- reads sequencing (10XWGS) technology aims to mitigate limitations by linking short-reads to the original larger fragment of DNA (~10kb). In this study, we performed an unbiased evaluation of these two technologies with different types and sizes of SVs and compared their performance. The SVs commonly identified by both the technologies were highly specific, while the validation rate dropped for uncommon SVs. Despite the technological advantage, a particularly high false discovery rate (FDR) was observed for SVs found only by 10XWGS without any significant improvement in sensitivity. We proposed a sensitive and specific statistical approach to improve SV predictions from both technologies and characterized SVs from MCF7 breast cancer cell line and a primary breast tumor with high precision.
Due to the limited benefit of 10XWGS for sensitivity, we trained a random forest classifier in FuseSV for accurate predictions only from cWGS sequencing data. FuseSV integrates SV predictions from multiple bioinformatics tools and mitigates high FDR of cWGS with a novel set of features derived from alignment of reads to the reference genome, biological mechanisms of SVs and breakpoints of SVs clustered together to consider complex genomic rearrangements (CGRs). The performance of FuseSV classifiers was superior to all individual bioinformatics tools as well as combined use with 10XWGS.
SVs whether simple or complex can form chimeric fusion transcripts (CMTs). CMTs can be predicted from RNA-sequencing (RNA-seq) data but include also transcripts that occur without underlying mutation and are also present in healthy tissues. Here we propose a novel pipeline, FUdGE, that predict three types of CMT directly from somatic SVs: These include direct fusion transcripts or classical fusion genes, transcripts with intron (IR) and intergenic region retained (INR). FUdGE allows independent confirmation of expressed CMTs from matched RNA-seq data. We validated the approach in the same MCF7 cell line and a primary breast tumor sample and investigate CMTs in a cohort of liposarcoma samples. Here we observed that the majority of confirmed SV driven CMTs were classical fusion genes with a much smaller number of IR and INR events.
Conclusively, FuseSV enables accurate prediction of somatic SVs in cancer using only cWGS. While FUdGE provides an RNA-seq independent strategy for direct prediction of CMTs formed due to somatic SV event. The respective expressed CMT candidates can be confirmed independently with RNA-seq data. This alternative approach only predicts tumor-specific somatic SV driven CMTs, which is advantageous for personalized immunotherapy interventions considering CMTs as neo-antigen candidates.XII, 91 Seiten ; Illustrationen, Diagramm
Integrative analysis of structural variations using short-reads and linked-reads yields highly specific and sensitive predictions.
No full textGenetic diseases are driven by aberrations of the human genome. Identification of such aberrations including structural variations (SVs) is key to our understanding. Conventional short-reads whole genome sequencing (cWGS) can identify SVs to base-pair resolution, but utilizes only short-range information and suffers from high false discovery rate (FDR). Linked-reads sequencing (10XWGS) utilizes long-range information by linkage of short-reads originating from the same large DNA molecule. This can mitigate alignment-based artefacts especially in repetitive regions and should enable better prediction of SVs. However, an unbiased evaluation of this technology is not available. In this study, we performed a comprehensive analysis of different types and sizes of SVs predicted by both the technologies and validated with an independent PCR based approach. The SVs commonly identified by both the technologies were highly specific, while validation rate dropped for uncommon events. A particularly high FDR was observed for SVs only found by 10XWGS. To improve FDR and sensitivity, statistical models for both the technologies were trained. Using our approach, we characterized SVs from the MCF7 cell line and a primary breast cancer tumor with high precision. This approach improves SV prediction and can therefore help in understanding the underlying genetics in various diseases
Statistical comparison of gene expression similarity (gene similarity scores).
No full textStatistical comparison of gene expression similarity (gene similarity scores).</p
The identified central hub nodes (TFs/miRNAs) and their enrichment analysis.
No full textThe identified central hub nodes (TFs/miRNAs) and their enrichment analysis.</p
The most significant TF-miRNA co-regulatory motifs for 4 hematopoietic lineages.
No full textTFs are represented by a turquoise triangle whereas miRNAs are shown as orange squares. Green circles denote the pluripotent genes whereas imprinted genes are colored in pink. Bold label nodes are supported by literature evidence.</p
Selected hematopoietic lineages and their stages of sequential cell development.
No full textSelected hematopoietic lineages and their stages of sequential cell development.</p
Functional homogeneity of the identified co-regulatory motifs.
No full textCumulative distribution of GO functional semantic scores of gene pairs of co-regulated genes in the examined motifs (red) versus randomly selected genes (black). The p-value was calculated using the Kolmogorov-Smirnov test.</p
Heatmaps of differentially expressed imprinted genes.
No full textThe order of genes is obtained by hierarchical clustering of three blood lineages (NK-cells, Monocytes, and Erythrocytes) based on the GSE34723 dataset. Gene clustering color coding is (blue) for paternally expressed genes, (red) for maternally expressed, (cyan) for pluripotent genes, and (orange) for hematopoietic genes. Shared genes between the pluripotent and hematopoietic gene sets are marked in black. Green spots represent downregulated genes, and red spots represent upregulated genes. The clustering reveals that for every lineage, there exist imprinted as well pluripotent and hematopoietic genes showing similar expression changes during cell development. The other three lineages (B-cells, T-cells, and granulocytes) are shown in the supplementary Figure D in S1 File.</p
Heatmaps showing transient changes in expression profiles.
No full textDifferent groups of ESC and hematopoietic cells (e.g stem cells, intermediate progenitors, and terminally differentiated blood cells) from the GSE10246 dataset for (left panel) imprinted genes, (middle panel) pluripotent genes and (right panel) hematopoietic genes were compared. Green spots represent downregulated genes, and red spots represent upregulated genes. The order of genes is obtained by hierarchical clustering, which shows three similar pattern classes between imprinted, pluripotent and hematopoietic genes.</p
