Bahman Afsari

Detection and localization of surgically resectable cancers with a multi-analyte blood test

SEEK and you may find cancer earlier Many cancers can be cured by surgery and/or systemic therapies when detected before they have metastasized. This clinical reality, coupled with the growing appreciation that cancers rapid genetic evolution limits its response to drugs, have fueled interest in methodologies for earlier detection of the disease. Cohen et al. developed a noninvasive blood test, called CancerSEEK that can detect eight common human cancer types (see the Perspective by Kalinich and Haber). The test assesses eight circulating protein biomarkers and tumor-specific mutations in circulating DNA. In a study of 1000 patients previously diagnosed with cancer and 850 healthy control individuals, CancerSEEK detected cancer with a sensitivity of 69 to 98% (depending on cancer type) and 99% specificity. Science, this issue p. 926; see also p. 866 A blood test that combines protein and DNA markers may allow earlier detection of eight common cancer types. Earlier detection is key to reducing cancer deaths. Here, we describe a blood test that can detect eight common cancer types through assessment of the levels of circulating proteins and mutations in cell-free DNA. We applied this test, called CancerSEEK, to 1005 patients with nonmetastatic, clinically detected cancers of the ovary, liver, stomach, pancreas, esophagus, colorectum, lung, or breast. CancerSEEK tests were positive in a median of 70% of the eight cancer types. The sensitivities ranged from 69 to 98% for the detection of five cancer types (ovary, liver, stomach, pancreas, and esophagus) for which there are no screening tests available for average-risk individuals. The specificity of CancerSEEK was greater than 99%: only 7 of 812 healthy controls scored positive. In addition, CancerSEEK localized the cancer to a small number of anatomic sites in a median of 83% of the patients.

Inferring causal molecular networks: empirical assessment through a community-based effort

It remains unclear whether causal, rather than merely correlational, relationships in molecular networks can be inferred in complex biological settings. Here we describe the HPN-DREAM network inference challenge, which focused on learning causal influences in signaling networks. We used phosphoprotein data from cancer cell lines as well as in silico data from a nonlinear dynamical model. Using the phosphoprotein data, we scored more than 2,000 networks submitted by challenge participants. The networks spanned 32 biological contexts and were scored in terms of causal validity with respect to unseen interventional data. A number of approaches were effective, and incorporating known biology was generally advantageous. Additional sub-challenges considered time-course prediction and visualization. Our results suggest that learning causal relationships may be feasible in complex settings such as disease states. Furthermore, our scoring approach provides a practical way to empirically assess inferred molecular networks in a causal sense.

A simple and reproducible breast cancer prognostic test

BackgroundA small number of prognostic and predictive tests based on gene expression are currently offered as reference laboratory tests. In contrast to such success stories, a number of flaws and errors have recently been identified in other genomic-based predictors and the success rate for developing clinically useful genomic signatures is low. These errors have led to widespread concerns about the protocols for conducting and reporting of computational research. As a result, a need has emerged for a template for reproducible development of genomic signatures that incorporates full transparency, data sharing and statistical robustness.ResultsHere we present the first fully reproducible analysis of the data used to train and test MammaPrint, an FDA-cleared prognostic test for breast cancer based on a 70-gene expression signature. We provide all the software and documentation necessary for researchers to build and evaluate genomic classifiers based on these data. As an example of the utility of this reproducible research resource, we develop a simple prognostic classifier that uses only 16 genes from the MammaPrint signature and is equally accurate in predicting 5-year disease free survival.ConclusionsOur study provides a prototypic example for reproducible development of computational algorithms for learning prognostic biomarkers in the era of personalized medicine.

Microarray Classification from Several Two-Gene Expression Comparisons

We describe our contribution to the ICMLA2008 ¿Automated Micro-Array Classification Challenge¿. The design of our classifier is motivated by the special scenario encountered in molecular cancer classification based on the mRNA concentrations provided by gene microarray data. Our classifier is rank-based; it only depends on expression comparisons among selected pairs of genes. Such comparisons are invariant to most of the transformations involved in preprocessing and normalization. Every pair of genes determines a binary classifier - choose the class for which the observed ordering is most likely. Pairs are scored by maximizing accuracy. In our k-TSP (k-disjoint Top Scoring Pairs) classifier, k disjoint pairs of genes are learned from training data; the discriminant function is simply the difference in the number of votes for the two classes. This rule involves exactly 2k genes, is readily interpretable, and provides some state-of-the-art results in cancer diagnosis and prognosis for small values of k, even k=1.

Evaluation of liquid from the Papanicolaou test and other liquid biopsies for the detection of endometrial and ovarian cancers

Endometrial and ovarian cancers can be detected through the analysis of DNA from Pap test fluids, intrauterine samples, and plasma. Brushing up on early cancer detection Despite the many recent advances in cancer diagnosis and treatment, ovarian cancer remains one of the most lethal malignancies, in part because there are no accurate screening methods for this disease and it is often diagnosed at a late stage. To develop a screening tool for ovarian and endometrial cancers, Wang et al. combined genetic analysis of fluids obtained through routine Papanicolau testing, normally done for cervical cancer, with analysis of tumor DNA circulating in the blood. The authors also used intrauterine sampling with Tao brushes to further increase the sensitivity of detection for the less accessible tumors. We report the detection of endometrial and ovarian cancers based on genetic analyses of DNA recovered from the fluids obtained during a routine Papanicolaou (Pap) test. The new test, called PapSEEK, incorporates assays for mutations in 18 genes as well as an assay for aneuploidy. In Pap brush samples from 382 endometrial cancer patients, 81% [95% confidence interval (CI), 77 to 85%] were positive, including 78% of patients with early-stage disease. The sensitivity in 245 ovarian cancer patients was 33% (95% CI, 27 to 39%), including 34% of patients with early-stage disease. In contrast, only 1.4% of 714 women without cancer had positive Pap brush samples (specificity, ~99%). Next, we showed that intrauterine sampling with a Tao brush increased the detection of malignancy over endocervical sampling with a Pap brush: 93% of 123 (95% CI, 87 to 97%) patients with endometrial cancer and 45% of 51 (95% CI, 31 to 60%) patients with ovarian cancer were positive, whereas none of the samples from 125 women without cancer were positive (specificity, 100%). Finally, in 83 ovarian cancer patients in whom plasma was available, circulating tumor DNA was found in 43% of patients (95% CI, 33 to 55%). When plasma and Pap brush samples were both tested, the sensitivity for ovarian cancer increased to 63% (95% CI, 51 to 73%). These results demonstrate the potential of mutation-based diagnostics to detect gynecologic cancers at a stage when they are more likely to be curable.

Integrated Analysis of Whole-Genome ChIP-Seq and RNA-Seq Data of Primary Head and Neck Tumor Samples Associates HPV Integration Sites with Open Chromatin Marks

Chromatin alterations mediate mutations and gene expression changes in cancer. Chromatin immunoprecipitation followed by sequencing (ChIP-Seq) has been utilized to study genome-wide chromatin structure in human cancer cell lines, yet numerous technical challenges limit comparable analyses in primary tumors. Here we have developed a new whole-genome analytic pipeline to optimize ChIP-Seq protocols on patient-derived xenografts from human papillomavirus-related (HPV+) head and neck squamous cell carcinoma (HNSCC) samples. We further associated chromatin aberrations with gene expression changes from a larger cohort of the tumor and normal samples with RNA-Seq data. We detect differential histone enrichment associated with tumor-specific gene expression variation, sites of HPV integration in the human genome, and HPV-associated histone enrichment sites upstream of cancer driver genes, which play central roles in cancer-associated pathways. These comprehensive analyses enable unprecedented characterization of the complex network of molecular changes resulting from chromatin alterations that drive HPV-related tumorigenesis. Cancer Res; 77(23); 6538-50. ©2017 AACR.

A Novel Functional Splice Variant of AKT3 Defined by Analysis of Alternative Splice Expression in HPV-Positive Oropharyngeal Cancers

The incidence of HPV-related oropharyngeal squamous cell carcinoma (OPSCC) has increased more than 200% in the past 20 years. Recent genetic sequencing efforts have elucidated relevant genes in head and neck cancer, but HPV-related tumors have consistently shown few DNA mutations. In this study, we sought to analyze alternative splicing events (ASE) that could alter gene function independent of mutations. To identify ASE unique to HPV-related tumors, RNA sequencing was performed on 46 HPV-positive OPSCC and 25 normal tissue samples. A novel algorithm using outlier statistics on RNA-sequencing junction expression identified 109 splicing events, which were confirmed in a validation set from The Cancer Genome Atlas. Because the most common type of splicing event identified was an alternative start site (39%), MBD-seq genome-wide CpG methylation data were analyzed for methylation alterations at promoter regions. ASE in six genes showed significant negative correlation between promoter methylation and expression of an alternative transcriptional start site, including AKT3 The novel AKT3 transcriptional variant and methylation changes were confirmed using qRT-PCR and qMSP methods. In vitro silencing of the novel AKT3 variant resulted in significant growth inhibition of multiple head and neck cell lines, an effect not observed with wild-type AKT3 knockdown. Analysis of ASE in HPV-related OPSCC identified multiple alterations likely involved in carcinogenesis, including a novel, functionally active transcriptional variant of AKT3 Our data indicate that ASEs represent a significant mechanism of oncogenesis with untapped potential for understanding complex genetic changes that result in the development of cancer. Cancer Res; 77(19); 5248-58. ©2017 AACR.

Non-invasive detection of urothelial cancer through the analysis of driver gene mutations and aneuploidy

Current non-invasive approaches for detection of urothelial cancers are suboptimal. We developed a test to detect urothelial neoplasms using DNA recovered from cells shed into urine. UroSEEK incorporates massive parallel sequencing assays for mutations in 11 genes and copy number changes on 39 chromosome arms. In 570 patients at risk for bladder cancer (BC), UroSEEK was positive in 83% of those who developed BC. Combined with cytology, UroSEEK detected 95% of patients who developed BC. Of 56 patients with upper tract urothelial cancer, 75% tested positive by UroSEEK, including 79% of those with non-invasive tumors. UroSEEK detected genetic abnormalities in 68% of urines obtained from BC patients under surveillance who demonstrated clinical evidence of recurrence. The advantages of UroSEEK over cytology were evident in low-grade BCs; UroSEEK detected 67% of cases whereas cytology detected none. These results establish the foundation for a new non-invasive approach for detection of urothelial cancer.

Splice Expression Variation Analysis (SEVA) for Differential Gene Isoform Usage in Cancer

Alternative splicing events (ASE) cause expression of a variable repertoire of potential protein products that are critical to carcinogenesis. Current methods to detect ASEs in tumor samples compare mean expression of gene isoforms relative to that of normal samples. However, these comparisons may not account for heterogeneous gene isoform usage in tumors. Therefore, we introduce Splice Expression Variability Analysis (SEVA) to detect differential splice variation, which accounts for tumor heterogeneity. This algorithm compares the degree of variability of junction expression profiles within a population of normal samples relative to that in tumor samples using a rank-based multivariate statistic that models the biological structure of ASEs. Simulated data show that SEVA is more robust to tumor heterogeneity and its candidates are more independent of differential expression than EBSeq and DiffSplice. SEVA analysis of head and neck tumors identified differential gene isoform usage robust in cross-study validation. Moreover, SEVA finds approximately hundreds of splice variant candidates, manageable for experimental validation in contrast to the thousands of candidates found with EBSeq or DiffSplice. Based on performance in both simulated and real data, SEVA is well suited for differential ASE analysis in RNA-sequencing data from heterogeneous primary tumor samples.

REVA: a rank-based multi-dimensional measure of correlation

The neighbors principle implicit in any machine learning algorithm says that samples with similar labels should be close to one another in feature space as well. For example, while tumors are heterogeneous, tumors that have similar genomics profiles can also be expected to have similar responses to a specific therapy. Simple correlation coefficients provide an effective way to determine whether this principle holds when features and labels are both scalar, but not when either is multivariate. A new class of generalized correlation coefficients based on inter-point distances addresses this need and is called “distance correlation”. There is only one rank-based distance correlation test available to date, and it is asymmetric in the samples, requiring that one sample be distinguished as a fixed point of reference. Therefore, we introduce a novel, nonparametric statistic, REVA, inspired by the Kendall rank correlation coefficient. We use U-statistic theory to derive the asymptotic distribution of the new correlation coefficient, developing additional large and finite sample properties along the way. To establish the admissibility of the REVA statistic, and explore the utility and limitations of our model, we compared it to the most widely used distance based correlation coefficient in a range of simulated conditions, demonstrating that REVA does not depend on an assumption of linearity, and is robust to high levels of noise, high dimensions, and the presence of outliers. We also present an application to real data, applying REVA to determine whether cancer cells with similar genetic profiles also respond similarly to a targeted therapeutic. Author summary Sometimes a simple question arises: how does the distance between two samples in multivariate space compare to another scalar value associated with each sample. Here, we propose theory for a nonparametric test to statistically test this association. This test is independent of the scale of the scalar data, and thus generalizable to any comparison of samples with both high-dimensional data and a scalar. We apply the resulting statistic, REVA, to problems in cancer biology motivated by the model that cancer cells with more similar gene expression profiles to one another can be expected to have a more similar response to therapy.