Unpaired data empowers association tests.

MOTIVATION: There is growing interest in the biomedical research community to incorporate retrospective data, available in healthcare systems, to shed light on associations between different biomarkers. Understanding the association between various types of biomedical data, such as genetic, blood biomarkers, imaging, etc. can provide a holistic understanding of human diseases. To formally test a hypothesized association between two types of data in Electronic Health Records (EHRs), one requires a substantial sample size with both data modalities to achieve a reasonable power. Current association test methods only allow using data from individuals who have both data modalities. Hence, researchers cannot take advantage of much larger EHR samples that includes individuals with at least one of the data types, which limits the power of the association test. RESULTS: We present a new method called the Semi-paired Association Test (SAT) that makes use of both paired and unpaired data. In contrast to classical approaches, incorporating unpaired data allows SAT to produce better control of false discovery and to improve the power of the association test. We study the properties of the new test theoretically and empirically, through a series of simulations and by applying our method on real studies in the context of Chronic Obstructive Pulmonary Disease. We are able to identify an association between the high-dimensional characterization of Computed Tomography chest images and several blood biomarkers as well as the expression of dozens of genes involved in the immune system. AVAILABILITY AND IMPLEMENTATION: Code is available on https://github.com/batmanlab/Semi-paired-Association-Test. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Low-Dimensional Density Ratio Estimation for Covariate Shift Correction.

Covariate shift is a prevalent setting for supervised learning in the wild when the training and test data are drawn from different time periods, different but related domains, or via different sampling strategies. This paper addresses a transfer learning setting, with covariate shift between source and target domains. Most existing methods for correcting covariate shift exploit density ratios of the features to reweight the source-domain data, and when the features are high-dimensional, the estimated density ratios may suffer large estimation variances, leading to poor prediction performance. In this work, we investigate the dependence of covariate shift correction performance on the dimensionality of the features, and propose a correction method that finds a low-dimensional representation of the features, which takes into account feature relevant to the target Y, and exploits the density ratio of this representation for importance reweighting. We discuss the factors affecting the performance of our method and demonstrate its capabilities on both pseudo-real and real-world data.

Data-Driven Approach to Multiple-Source Domain Adaptation.

A key problem in domain adaptation is determining what to transfer across different domains. We propose a data-driven method to represent these changes across multiple source domains and perform unsupervised domain adaptation. We assume that the joint distributions follow a specific generating process and have a small number of identifiable changing parameters, and develop a data-driven method to identify the changing parameters by learning low-dimensional representations of the changing class-conditional distributions across multiple source domains. The learned low-dimensional representations enable us to reconstruct the target-domain joint distribution from unlabeled target-domain data, and further enable predicting the labels in the target domain. We demonstrate the efficacy of this method by conducting experiments on synthetic and real datasets.

Reconstructing cancer drug response networks using multitask learning.

BACKGROUND: Translating in vitro results to clinical tests is a major challenge in systems biology. Here we present a new Multi-Task learning framework which integrates thousands of cell line expression experiments to reconstruct drug specific response networks in cancer. RESULTS: The reconstructed networks correctly identify several shared key proteins and pathways while simultaneously highlighting many cell type specific proteins. We used top proteins from each drug network to predict survival for patients prescribed the drug. CONCLUSIONS: Predictions based on proteins from the in-vitro derived networks significantly outperformed predictions based on known cancer genes indicating that Multi-Task learning can indeed identify accurate drug response networks.

Recurrent and functional regulatory mutations in breast cancer.

Genomic analysis of tumours has led to the identification of hundreds of cancer genes on the basis of the presence of mutations in protein-coding regions. By contrast, much less is known about cancer-causing mutations in non-coding regions. Here we perform deep sequencing in 360 primary breast cancers and develop computational methods to identify significantly mutated promoters. Clear signals are found in the promoters of three genes. FOXA1, a known driver of hormone-receptor positive breast cancer, harbours a mutational hotspot in its promoter leading to overexpression through increased E2F binding. RMRP and NEAT1, two non-coding RNA genes, carry mutations that affect protein binding to their promoters and alter expression levels. Our study shows that promoter regions harbour recurrent mutations in cancer with functional consequences and that the mutations occur at similar frequencies as in coding regions. Power analyses indicate that more such regions remain to be discovered through deep sequencing of adequately sized cohorts of patients.

Mutational Strand Asymmetries in Cancer Genomes Reveal Mechanisms of DNA Damage and Repair.

Mutational processes constantly shape the somatic genome, leading to immunity, aging, cancer, and other diseases. When cancer is the outcome, we are afforded a glimpse into these processes by the clonal expansion of the malignant cell. Here, we characterize a less explored layer of the mutational landscape of cancer: mutational asymmetries between the two DNA strands. Analyzing whole-genome sequences of 590 tumors from 14 different cancer types, we reveal widespread asymmetries across mutagenic processes, with transcriptional ("T-class") asymmetry dominating UV-, smoking-, and liver-cancer-associated mutations and replicative ("R-class") asymmetry dominating POLE-, APOBEC-, and MSI-associated mutations. We report a striking phenomenon of transcription-coupled damage (TCD) on the non-transcribed DNA strand and provide evidence that APOBEC mutagenesis occurs on the lagging-strand template during DNA replication. As more genomes are sequenced, studying and classifying their asymmetries will illuminate the underlying biological mechanisms of DNA damage and repair.

Paired exome analysis of Barrett's esophagus and adenocarcinoma.

Barrett's esophagus is thought to progress to esophageal adenocarcinoma (EAC) through a stepwise progression with loss of CDKN2A followed by TP53 inactivation and aneuploidy. Here we present whole-exome sequencing from 25 pairs of EAC and Barrett's esophagus and from 5 patients whose Barrett's esophagus and tumor were extensively sampled. Our analysis showed that oncogene amplification typically occurred as a late event and that TP53 mutations often occurred early in Barrett's esophagus progression, including in non-dysplastic epithelium. Reanalysis of additional EAC exome data showed that the majority (62.5%) of EACs emerged following genome doubling and that tumors with genomic doubling had different patterns of genomic alterations, with more frequent oncogenic amplification and less frequent inactivation of tumor suppressors, including CDKN2A. These data suggest that many EACs emerge not through the gradual accumulation of tumor-suppressor alterations but rather through a more direct path whereby a TP53-mutant cell undergoes genome doubling, followed by the acquisition of oncogenic amplifications.

Low incidence of complications after cephalic vein cutdown for pacemaker lead implantation in children weighing less than 10 kilograms: A single-center experience with long-term follow-up.

BACKGROUND: Only a few studies on the cephalic vein cutdown technique for pacemaker lead implantation in children weighing ≤10 kg have been reported even though the procedure is widely accepted in adults. OBJECTIVE: The purpose of this study was to prove that cephalic vein cutdown for pacemaker lead implantation is a reliable technique with a low incidence of complications in children weighing ≤10 kg. METHODS: The study included 44 children weighing ≤10 kg with an endocardial pacemaker. Cephalic, subclavian, and axillary vein diameters were measured by ultrasound before implantation. The measured diameters were used to select either an endocardial or epicardial surgical technique. Regular 6-month follow-up visits included pacemaker interrogation and clinical and ultrasound examinations. RESULTS: Two dual-chamber and 42 single-chamber pacemakers were implanted. Mean weight at implantation was 6.24 kg (range 2.25-10.40 kg), and mean age was 11.4 months (range 1 day-47 months). In 40 children (90.1%), the ventricular leads were implanted using the cephalic vein cutdown technique, and implantation was accomplished via the prepared right external jugular vein in 4 of the children (9.9%). The atrial leads were implanted using axillary vein puncture and external jugular vein preparations. Mean follow-up was 8.9 years (range 0-20.9 years). Only 1 pacemaker-related complication was detected (a lead fracture near the connector that was successfully resolved using a lead repair kit). CONCLUSION: The cephalic vein cutdown technique is feasible and reliable in children weighing ≤10 kg, which justifies the application of additional surgical effort in the treatment of these small patients.

Integrative and comparative genomic analysis of HPV-positive and HPV-negative head and neck squamous cell carcinomas.

PURPOSE: The genetic differences between human papilloma virus (HPV)-positive and -negative head and neck squamous cell carcinomas (HNSCC) remain largely unknown. To identify differential biology and novel therapeutic targets for both entities, we determined mutations and copy-number aberrations in a large cohort of locoregionally advanced HNSCC. EXPERIMENTAL DESIGN: We performed massively parallel sequencing of 617 cancer-associated genes in 120 matched tumor/normal samples (42.5% HPV-positive). Mutations and copy-number aberrations were determined and results validated with a secondary method. RESULTS: The overall mutational burden in HPV-negative and HPV-positive HNSCC was similar with an average of 15.2 versus 14.4 somatic exonic mutations in the targeted cancer-associated genes. HPV-negative tumors showed a mutational spectrum concordant with published lung squamous cell carcinoma analyses with enrichment for mutations in TP53, CDKN2A, MLL2, CUL3, NSD1, PIK3CA, and NOTCH genes. HPV-positive tumors showed unique mutations in DDX3X, FGFR2/3 and aberrations in PIK3CA, KRAS, MLL2/3, and NOTCH1 were enriched in HPV-positive tumors. Currently targetable genomic alterations were identified in FGFR1, DDR2, EGFR, FGFR2/3, EPHA2, and PIK3CA. EGFR, CCND1, and FGFR1 amplifications occurred in HPV-negative tumors, whereas 17.6% of HPV-positive tumors harbored mutations in fibroblast growth factor receptor genes (FGFR2/3), including six recurrent FGFR3 S249C mutations. HPV-positive tumors showed a 5.8% incidence of KRAS mutations, and DNA-repair gene aberrations, including 7.8% BRCA1/2 mutations, were identified. CONCLUSIONS: The mutational makeup of HPV-positive and HPV-negative HNSCC differs significantly, including targetable genes. HNSCC harbors multiple therapeutically important genetic aberrations, including frequent aberrations in the FGFR and PI3K pathway genes. See related commentary by Krigsfeld and Chung, p. 495.

TET2 mutations predict response to hypomethylating agents in myelodysplastic syndrome patients.

Only a minority of myelodysplastic syndrome (MDS) patients respond to hypomethylating agents (HMAs), but strong predictors of response are unknown. We sequenced 40 recurrently mutated myeloid malignancy genes in tumor DNA from 213 MDS patients collected before treatment with azacitidine (AZA) or decitabine (DEC). Mutations were examined for association with response and overall survival. The overall response rate of 47% was not different between agents. Clonal TET2 mutations predicted response (odds ratio [OR] 1.99, P = .036) when subclones unlikely to be detected by Sanger sequencing (allele fraction <10%) were treated as wild-type (WT). Response rates were highest in the subset of TET2 mutant patients without clonal ASXL1 mutations (OR 3.65, P = .009). Mutations of TP53 (hazard ratio [HR] 2.01, P = .002) and PTPN11 (HR 3.26, P = .006) were associated with shorter overall survival but not drug response. Murine-competitive bone marrow transplantation followed by treatment with AZA demonstrated that Tet2-null cells have an engraftment advantage over Tet2-WT cells. AZA significantly decreased this advantage for Tet2-null cells (P = .002) but not Tet2-WT cells (P = .212). Overall, Tet2 loss appears to sensitize cells to treatment with AZA in vivo, and TET2 mutations can identify patients more likely to respond to HMAs.

The genomic landscape of pediatric Ewing sarcoma.

UNLABELLED: Pediatric Ewing sarcoma is characterized by the expression of chimeric fusions of EWS and ETS family transcription factors, representing a paradigm for studying cancers driven by transcription factor rearrangements. In this study, we describe the somatic landscape of pediatric Ewing sarcoma. These tumors are among the most genetically normal cancers characterized to date, with only EWS-ETS rearrangements identified in the majority of tumors. STAG2 loss, however, is present in more than 15% of Ewing sarcoma tumors; occurs by point mutation, rearrangement, and likely nongenetic mechanisms; and is associated with disease dissemination. Perhaps the most striking finding is the paucity of mutations in immediately targetable signal transduction pathways, highlighting the need for new therapeutic approaches to target EWS-ETS fusions in this disease. SIGNIFICANCE: We performed next-generation sequencing of Ewing sarcoma, a pediatric cancer involving bone, characterized by expression of EWS-ETS fusions. We found remarkably few mutations. However, we discovered that loss of STAG2 expression occurs in 15% of tumors and is associated with metastatic disease, suggesting a potential genetic vulnerability in Ewing sarcoma.

Somatic mutations predict poor outcome in patients with myelodysplastic syndrome after hematopoietic stem-cell transplantation.

PURPOSE: Recurrently mutated genes in myelodysplastic syndrome (MDS) are pathogenic drivers and powerfully associated with clinical phenotype and prognosis. Whether these types of mutations predict outcome after allogeneic hematopoietic stem-cell transplantation (HSCT) in patients with MDS is not known. PATIENTS AND METHODS: We used massively parallel sequencing to examine tumor samples collected from 87 patients with MDS before HSCT for coding mutations in 40 recurrently mutated MDS genes. RESULTS: Mutations were identified in 92% of patients, most frequently in the ASXL1 (29%), TP53 (21%), DNMT3A (18%), and RUNX1 (16%) genes. In univariable analyses, only TP53 mutations were associated with shorter overall (OS; hazard ratio [HR], 3.74; P < .001) and progression-free survival (HR, 3.97; P < .001). After adjustment for clinical variables associated with these end points, mutations in TP53 (HR, 2.30; P = .027), TET2 (HR, 2.40; P = .033), and DNMT3A (HR, 2.08; P = .049) were associated with decreased OS. In multivariable analysis including clinical variables, complex karyotype status, and candidate genes, mutations in TP53 (HR, 4.22; P ≤ .001) and TET2 (HR, 1.68; P = .037) were each independently associated with shorter OS. Nearly one half of patients (46%) carried a mutation in TP53, DNMT3A, or TET2 and accounted for 64% of deaths. Three-year OS in patients without these mutations was 59% (95% CI, 43% to 72%), versus 19% (95% CI, 9% to 33%) in patients with these mutations. CONCLUSION: Mutations in TP53, TET2, or DNMT3A identify patients with MDS with shorter OS after HSCT.

Cancer therapy. Ex vivo culture of circulating breast tumor cells for individualized testing of drug susceptibility.

Circulating tumor cells (CTCs) are present at low concentrations in the peripheral blood of patients with solid tumors. It has been proposed that the isolation, ex vivo culture, and characterization of CTCs may provide an opportunity to noninvasively monitor the changing patterns of drug susceptibility in individual patients as their tumors acquire new mutations. In a proof-of-concept study, we established CTC cultures from six patients with estrogen receptor-positive breast cancer. Three of five CTC lines tested were tumorigenic in mice. Genome sequencing of the CTC lines revealed preexisting mutations in the PIK3CA gene and newly acquired mutations in the estrogen receptor gene (ESR1), PIK3CA gene, and fibroblast growth factor receptor gene (FGFR2), among others. Drug sensitivity testing of CTC lines with multiple mutations revealed potential new therapeutic targets. With optimization of CTC culture conditions, this strategy may help identify the best therapies for individual cancer patients over the course of their disease.

Whole-exome sequencing and clinical interpretation of formalin-fixed, paraffin-embedded tumor samples to guide precision cancer medicine.

Translating whole-exome sequencing (WES) for prospective clinical use may have an impact on the care of patients with cancer; however, multiple innovations are necessary for clinical implementation. These include rapid and robust WES of DNA derived from formalin-fixed, paraffin-embedded tumor tissue, analytical output similar to data from frozen samples and clinical interpretation of WES data for prospective use. Here, we describe a prospective clinical WES platform for archival formalin-fixed, paraffin-embedded tumor samples. The platform employs computational methods for effective clinical analysis and interpretation of WES data. When applied retrospectively to 511 exomes, the interpretative framework revealed a 'long tail' of somatic alterations in clinically important genes. Prospective application of this approach identified clinically relevant alterations in 15 out of 16 patients. In one patient, previously undetected findings guided clinical trial enrollment, leading to an objective clinical response. Overall, this methodology may inform the widespread implementation of precision cancer medicine.

Somatic mutation as a mechanism of Wnt/ß-catenin pathway activation in CLL.

One major goal of cancer genome sequencing is to identify key genes and pathways that drive tumor pathogenesis. Although many studies have identified candidate driver genes based on recurrence of mutations in individual genes, subsets of genes with nonrecurrent mutations may also be defined as putative drivers if they affect a single biological pathway. In this fashion, we previously identified Wnt signaling as significantly mutated through large-scale massively parallel DNA sequencing of chronic lymphocytic leukemia (CLL). Here, we use a novel method of biomolecule delivery, vertical silicon nanowires, to efficiently introduce small interfering RNAs into CLL cells, and interrogate the effects of 8 of 15 mutated Wnt pathway members identified across 91 CLLs. In HEK293T cells, mutations in 2 genes did not generate functional changes, 3 led to dysregulated pathway activation, and 3 led to further activation or loss of repression of pathway activation. Silencing 4 of 8 mutated genes in CLL samples harboring the mutated alleles resulted in reduced viability compared with leukemia samples with wild-type alleles. We demonstrate that somatic mutations in CLL can generate dependence on this pathway for survival. These findings support the notion that nonrecurrent mutations at different nodes of the Wnt pathway can contribute to leukemogenesis.

Widespread genetic heterogeneity in multiple myeloma: implications for targeted therapy.

We performed massively parallel sequencing of paired tumor/normal samples from 203 multiple myeloma (MM) patients and identified significantly mutated genes and copy number alterations and discovered putative tumor suppressor genes by determining homozygous deletions and loss of heterozygosity. We observed frequent mutations in KRAS (particularly in previously treated patients), NRAS, BRAF, FAM46C, TP53, and DIS3 (particularly in nonhyperdiploid MM). Mutations were often present in subclonal populations, and multiple mutations within the same pathway (e.g., KRAS, NRAS, and BRAF) were observed in the same patient. In vitro modeling predicts only partial treatment efficacy of targeting subclonal mutations, and even growth promotion of nonmutated subclones in some cases. These results emphasize the importance of heterogeneity analysis for treatment decisions.

Discovery and saturation analysis of cancer genes across 21 tumour types.

Although a few cancer genes are mutated in a high proportion of tumours of a given type (>20%), most are mutated at intermediate frequencies (2-20%). To explore the feasibility of creating a comprehensive catalogue of cancer genes, we analysed somatic point mutations in exome sequences from 4,742 human cancers and their matched normal-tissue samples across 21 cancer types. We found that large-scale genomic analysis can identify nearly all known cancer genes in these tumour types. Our analysis also identified 33 genes that were not previously known to be significantly mutated in cancer, including genes related to proliferation, apoptosis, genome stability, chromatin regulation, immune evasion, RNA processing and protein homeostasis. Down-sampling analysis indicates that larger sample sizes will reveal many more genes mutated at clinically important frequencies. We estimate that near-saturation may be achieved with 600-5,000 samples per tumour type, depending on background mutation frequency. The results may help to guide the next stage of cancer genomics.

Reduced local mutation density in regulatory DNA of cancer genomes is linked to DNA repair.

Carcinogenesis and neoplastic progression are mediated by the accumulation of somatic mutations. Here we report that the local density of somatic mutations in cancer genomes is highly reduced specifically in accessible regulatory DNA defined by DNase I hypersensitive sites. This reduction is independent of any known factors influencing somatic mutation density and is observed in diverse cancer types, suggesting a general mechanism. By analyzing individual cancer genomes, we show that the reduced local mutation density within regulatory DNA is linked to intact global genome repair machinery, with nearly complete abrogation of the hypomutation phenomenon in individual cancers that possess mutations in components of the nucleotide excision repair system. Together, our results connect chromatin structure, gene regulation and cancer-associated somatic mutation.

Integrative and comparative genomic analysis of lung squamous cell carcinomas in East Asian patients.

PURPOSE: Lung squamous cell carcinoma (SCC) is the second most prevalent type of lung cancer. Currently, no targeted therapeutics are approved for treatment of this cancer, largely because of a lack of systematic understanding of the molecular pathogenesis of the disease. To identify therapeutic targets and perform comparative analyses of lung SCC, we probed somatic genome alterations of lung SCC by using samples from Korean patients. PATIENTS AND METHODS: We performed whole-exome sequencing of DNA from 104 lung SCC samples from Korean patients and matched normal DNA. In addition, copy-number analysis and transcriptome analysis were conducted for a subset of these samples. Clinical association with cancer-specific somatic alterations was investigated. RESULTS: This cancer cohort is characterized by a high mutational burden with an average of 261 somatic exonic mutations per tumor and a mutational spectrum showing a signature of exposure to cigarette smoke. Seven genes demonstrated statistical enrichment for mutation: TP53, RB1, PTEN, NFE2L2, KEAP1, MLL2, and PIK3CA). Comparative analysis between Korean and North American lung SCC samples demonstrated a similar spectrum of alterations in these two populations in contrast to the differences seen in lung adenocarcinoma. We also uncovered recurrent occurrence of therapeutically actionable FGFR3-TACC3 fusion in lung SCC. CONCLUSION: These findings provide new steps toward the identification of genomic target candidates for precision medicine in lung SCC, a disease with significant unmet medical needs.

An APOBEC cytidine deaminase mutagenesis pattern is widespread in human cancers.

Recent studies indicate that a subclass of APOBEC cytidine deaminases, which convert cytosine to uracil during RNA editing and retrovirus or retrotransposon restriction, may induce mutation clusters in human tumors. We show here that throughout cancer genomes APOBEC-mediated mutagenesis is pervasive and correlates with APOBEC mRNA levels. Mutation clusters in whole-genome and exome data sets conformed to the stringent criteria indicative of an APOBEC mutation pattern. Applying these criteria to 954,247 mutations in 2,680 exomes from 14 cancer types, mostly from The Cancer Genome Atlas (TCGA), showed a significant presence of the APOBEC mutation pattern in bladder, cervical, breast, head and neck, and lung cancers, reaching 68% of all mutations in some samples. Within breast cancer, the HER2-enriched subtype was clearly enriched for tumors with the APOBEC mutation pattern, suggesting that this type of mutagenesis is functionally linked with cancer development. The APOBEC mutation pattern also extended to cancer-associated genes, implying that ubiquitous APOBEC-mediated mutagenesis is carcinogenic.