A machine learning approach to optimizing cell-free DNA sequencing panels: with an application to prostate cancer.

BACKGROUND: Cell-free DNA's (cfDNA) use as a biomarker in cancer is challenging due to genetic heterogeneity of malignancies and rarity of tumor-derived molecules. Here we describe and demonstrate a novel machine-learning guided panel design strategy for improving the detection of tumor variants in cfDNA. Using this approach, we first generated a model to classify and score candidate variants for inclusion on a prostate cancer targeted sequencing panel. We then used this panel to screen tumor variants from prostate cancer patients with localized disease in both in silico and hybrid capture settings. METHODS: Whole Genome Sequence (WGS) data from 550 prostate tumors was analyzed to build a targeted sequencing panel of single point and small (< 200 bp) indel mutations, which was subsequently screened in silico against prostate tumor sequences from 5 patients to assess performance against commonly used alternative panel designs. The panel's ability to detect tumor-derived cfDNA variants was then assessed using prospectively collected cfDNA and tumor foci from a test set 18 prostate cancer patients with localized disease undergoing radical proctectomy. RESULTS: The panel generated from this approach identified as top candidates mutations in known driver genes (e.g. HRAS) and prostate cancer related transcription factor binding sites (e.g. MYC, AR). It outperformed two commonly used designs in detecting somatic mutations found in the cfDNA of 5 prostate cancer patients when analyzed in an in silico setting. Additionally, hybrid capture and 2500X sequencing of cfDNA molecules using the panel resulted in detection of tumor variants in all 18 patients of a test set, where 15 of the 18 patients had detected variants found in multiple foci. CONCLUSION: Machine learning-prioritized targeted sequencing panels may prove useful for broad and sensitive variant detection in the cfDNA of heterogeneous diseases. This strategy has implications for disease detection and monitoring when applied to the cfDNA isolated from prostate cancer patients.

Orchid: a novel management, annotation and machine learning framework for analyzing cancer mutations.

Motivation: As whole-genome tumor sequence and biological annotation datasets grow in size, number and content, there is an increasing basic science and clinical need for efficient and accurate data management and analysis software. With the emergence of increasingly sophisticated data stores, execution environments and machine learning algorithms, there is also a need for the integration of functionality across frameworks. Results: We present orchid, a python based software package for the management, annotation and machine learning of cancer mutations. Building on technologies of parallel workflow execution, in-memory database storage and machine learning analytics, orchid efficiently handles millions of mutations and hundreds of features in an easy-to-use manner. We describe the implementation of orchid and demonstrate its ability to distinguish tissue of origin in 12 tumor types based on 339 features using a random forest classifier. Availability and implementation: Orchid and our annotated tumor mutation database are freely available at https://github.com/wittelab/orchid. Software is implemented in python 2.7, and makes use of MySQL or MemSQL databases. Groovy 2.4.5 is optionally required for parallel workflow execution. Contact: JWitte@ucsf.edu. Supplementary information: Supplementary data are available at Bioinformatics online.

Mouse genome-wide association study identifies polymorphisms on chromosomes 4, 11, and 15 for age-related cardiac fibrosis.

Dystrophic cardiac calcinosis (DCC), also called epicardial and myocardial fibrosis and mineralization, has been detected in mice of a number of laboratory inbred strains, most commonly C3H/HeJ and DBA/2J. In previous mouse breeding studies between these DCC susceptible and the DCC-resistant strain C57BL/6J, 4 genetic loci harboring genes involved in DCC inheritance were identified and subsequently termed Dyscalc loci 1 through 4. Here, we report susceptibility to cardiac fibrosis, a sub-phenotype of DCC, at 12 and 20 months of age and close to natural death in a survey of 28 inbred mouse strains. Eight strains showed cardiac fibrosis with highest frequency and severity in the moribund mice. Using genotype and phenotype information of the 28 investigated strains, we performed genome-wide association studies (GWAS) and identified the most significant associations on chromosome (Chr) 15 at 72 million base pairs (Mb) (P < 10(-13)) and Chr 4 at 122 Mb (P < 10(-11)) and 134 Mb (P < 10(-7)). At the Chr 15 locus, Col22a1 and Kcnk9 were identified. Both have been reported to be morphologically and functionally important in the heart muscle. The strongest Chr 4 associations were located approximately 6 Mb away from the Dyscalc 2 quantitative trait locus peak within the boundaries of the Extl1 gene and in close proximity to the Trim63 and Cap1 genes. In addition, a single-nucleotide polymorphism association was found on chromosome 11. This study provides evidence for more than the previously reported 4 genetic loci determining cardiac fibrosis and DCC. The study also highlights the power of GWAS in the mouse for dissecting complex genetic traits.

A major X-linked locus affects kidney function in mice.

Chronic kidney disease is a common disease with increasing prevalence in the western population. One common reason for chronic kidney failure is diabetic nephropathy. Diabetic nephropathy and hyperglycemia are characteristics of the mouse inbred strain KK/HlJ, which is predominantly used as a model for metabolic syndrome due to its inherited glucose intolerance and insulin resistance. We used KK/HlJ, an albuminuria-sensitive strain, and C57BL/6J, an albuminuria-resistant strain, to perform a quantitative trait locus (QTL) cross to identify the genetic basis for chronic kidney failure. Albumin-creatinine ratio (ACR) was measured in 130 F2 male offspring. One significant QTL was identified on chromosome (Chr) X and four suggestive QTL were found on Chrs 6, 7, 12, and 13. Narrowing of the QTL region was focused on the X-linked QTL and performed by incorporating genotype and expression analyses for genes located in the region. From the 485 genes identified in the X-linked QTL region, a few candidate genes were identified using a combination of bioinformatic evidence based on genomic comparison of the parental strains and known function in urine homeostasis. Finally, this study demonstrates the significance of the X chromosome in the genetic determination of albuminuria.

Early gene expression differences in inbred mouse strains with susceptibility to pulmonary adenomas.

Lung cancer is the most common cause of cancer-related deaths in both men and women, and effective preventatives are rare due to the difficulty of early detection. Specific gene expression signatures have been identified in individuals that already developed lung cancer. To identify if gene expression differences could be detected in individuals before the onset of the disease, we obtained lung tissues for microarray analysis from young, healthy mice of 9 inbred strains with known differences in their susceptibility to spontaneous pulmonary adenomas when aged. We found that the most common differentially expressed genes among all possible 36 strain comparisons showed significant associations with cancer- and inflammation-related processes. Significant expression differences between susceptible and resistant strains were detected for Aldh3a1, Cxcr1 and 7, Dpt, and Nptx1-genes with known cancer-related functions, and Cd209, Cxcr1 and 7, and Plag2g1b-genes with known inflammatory-related functions. Whereas Aldh3a1, Cd209, Dpt, and Pla2g1b had increased expression, Cxcr1 and 7, and Nptx1 had decreased expression in strains susceptible to pulmonary adenomas. Thus, our study shows that expression differences between susceptible and resistant strains can be detected in young and healthy mice without manifestation of pulmonary adenomas and, thus, may provide an opportunity of early detection. Finally, the identified genes have previously been reported for human non-small cell lung cancer suggesting that molecular pathways may be shared between these two cancer types.

Automated measurement of zebrafish larval movement.

The zebrafish is a powerful vertebrate model that is readily amenable to genetic, pharmacological and environmental manipulations to elucidate the molecular and cellular basis of movement and behaviour. We report software enabling automated analysis of zebrafish movement from video recordings captured with cameras ranging from a basic camcorder to more specialized equipment. The software, which is provided as open-source MATLAB functions, can be freely modified and distributed, and is compatible with multiwell plates under a wide range of experimental conditions. Automated measurement of zebrafish movement using this technique will be useful for multiple applications in neuroscience, pharmacology and neuropsychiatry.

Evaluation of spontaneous propulsive movement as a screening tool to detect rescue of Parkinsonism phenotypes in zebrafish models.

Zebrafish models of human neuropsychiatric diseases offer opportunities to identify novel therapeutic targets and treatments through phenotype-based genetic or chemical modifier screens. In order to develop an assay to detect rescue of zebrafish models of Parkinsonism, we characterized spontaneous zebrafish larval motor behavior from 3 to 9 days post fertilization in a microtiter plate format suitable for screening, and clarified the role of dopaminergic signaling in its regulation. The proportion of time that larvae spent moving increased progressively between 3 and 9 dpf, whereas their active velocity decreased between 5 and 6 dpf as sporadic burst movements gave way to a more mature beat-and-glide pattern. Spontaneous movement varied between larvae and during the course of recordings as a result of intrinsic larval factors including genetic background. Variability decreased with age, such that small differences between groups of larvae exposed to different experimental conditions could be detected robustly by 6 to 7 dpf. Suppression of endogenous dopaminergic signaling by exposure to MPP(+), haloperidol or chlorpromazine reduced mean velocity by decreasing the frequency with which spontaneous movements were initiated, but did not alter active velocity. The variability of mean velocity assays could be reduced by analyzing groups of larvae for each data point, yielding acceptable screening window coefficients; the sample size required in each group was determined by the magnitude of the motor phenotype in different models. For chlorpromazine exposure, samples of four larvae allowed robust separation of treated and untreated data points (Z=0.42), whereas the milder impairment provoked by MPP(+) necessitated groups of eight larvae in order to provide a useful discovery assay (Z=0.13). Quantification of spontaneous larval movement offers a simple method to determine functional integrity of motor systems, and may be a useful tool to isolate novel molecular modulators of Parkinsonism phenotypes.

Cell-free DNA concentration and fragment size as a biomarker for prostate cancer.

Prostate cancer is the most commonly diagnosed neoplasm in American men. Although existing biomarkers may detect localized prostate cancer, additional strategies are necessary for improving detection and identifying aggressive disease that may require further intervention. One promising, minimally invasive biomarker is cell-free DNA (cfDNA), which consist of short DNA fragments released into circulation by dying or lysed cells that may reflect underlying cancer. Here we investigated whether differences in cfDNA concentration and cfDNA fragment size could improve the sensitivity for detecting more advanced and aggressive prostate cancer. This study included 268 individuals: 34 healthy controls, 112 men with localized prostate cancer who underwent radical prostatectomy (RP), and 122 men with metastatic castration-resistant prostate cancer (mCRPC). Plasma cfDNA concentration and fragment size were quantified with the Qubit 3.0 and the 2100 Bioanalyzer. The potential relationship between cfDNA concentration or fragment size and localized or mCRPC prostate cancer was evaluated with descriptive statistics, logistic regression, and area under the curve analysis with cross-validation. Plasma cfDNA concentrations were elevated in mCRPC patients in comparison to localized disease (OR5ng/mL = 1.34, P = 0.027) or to being a control (OR5ng/mL = 1.69, P = 0.034). Decreased average fragment size was associated with an increased risk of localized disease compared to controls (OR5bp = 0.77, P = 0.0008). This study suggests that while cfDNA concentration can identify mCRPC patients, it is unable to distinguish between healthy individuals and patients with localized prostate cancer. In addition to PSA, average cfDNA fragment size may be an alternative that can differentiate between healthy individuals and those with localized disease, but the low sensitivity and specificity results in an imperfect diagnostic marker. While quantification of cfDNA may provide a quick, cost-effective approach to help guide treatment decisions in advanced disease, its use is limited in the setting of localized prostate cancer.

Cell-Free DNA Detection of Tumor Mutations in Heterogeneous, Localized Prostate Cancer Via Targeted, Multiregion Sequencing.

Cell-free DNA (cfDNA) may allow for minimally invasive identification of biologically relevant genomic alterations and genetically distinct tumor subclones. Although existing biomarkers may detect localized prostate cancer, additional strategies interrogating genomic heterogeneity are necessary for identifying and monitoring aggressive disease. In this study, we aimed to evaluate whether circulating tumor DNA can detect genomic alterations present in multiple regions of localized prostate tumor tissue. METHODS: Low-pass whole-genome and targeted sequencing with a machine-learning guided 2.5-Mb targeted panel were used to identify single nucleotide variants, small insertions and deletions (indels), and copy-number alterations in cfDNA. The majority of this study focuses on the subset of 21 patients with localized disease, although 45 total individuals were evaluated, including 15 healthy controls and nine men with metastatic castration-resistant prostate cancer. Plasma cfDNA was barcoded with duplex unique molecular identifiers. For localized cases, matched tumor tissue was collected from multiple regions (one to nine samples per patient) for comparison. RESULTS: Somatic tumor variants present in heterogeneous tumor foci from patients with localized disease were detected in cfDNA, and cfDNA mutational burden was found to track with disease severity. Somatic tissue alterations were identified in cfDNA, including nonsynonymous variants in FOXA1, PTEN, MED12, and ATM. Detection of these overlapping variants was associated with seminal vesicle invasion (P = .019) and with the number of variants initially found in the matched tumor tissue samples (P = .0005). CONCLUSION: Our findings demonstrate the potential of targeted cfDNA sequencing to detect somatic tissue alterations in heterogeneous, localized prostate cancer, especially in a setting where matched tumor tissue may be unavailable (ie, active surveillance or treatment monitoring).

A Large-Scale Association Study Detects Novel Rare Variants, Risk Genes, Functional Elements, and Polygenic Architecture of Prostate Cancer Susceptibility.

To identify rare variants associated with prostate cancer susceptibility and better characterize the mechanisms and cumulative disease risk associated with common risk variants, we conducted an integrated study of prostate cancer genetic etiology in two cohorts using custom genotyping microarrays, large imputation reference panels, and functional annotation approaches. Specifically, 11,984 men (6,196 prostate cancer cases and 5,788 controls) of European ancestry from Northern California Kaiser Permanente were genotyped and meta-analyzed with 196,269 men of European ancestry (7,917 prostate cancer cases and 188,352 controls) from the UK Biobank. Three novel loci, including two rare variants (European ancestry minor allele frequency < 0.01, at 3p21.31 and 8p12), were significant genome wide in a meta-analysis. Gene-based rare variant tests implicated a known prostate cancer gene (HOXB13), as well as a novel candidate gene (ILDR1), which encodes a receptor highly expressed in prostate tissue and is related to the B7/CD28 family of T-cell immune checkpoint markers. Haplotypic patterns of long-range linkage disequilibrium were observed for rare genetic variants at HOXB13 and other loci, reflecting their evolutionary history. In addition, a polygenic risk score (PRS) of 188 prostate cancer variants was strongly associated with risk (90th vs. 40th-60th percentile OR = 2.62, P = 2.55 × 10-191). Many of the 188 variants exhibited functional signatures of gene expression regulation or transcription factor binding, including a 6-fold difference in log-probability of androgen receptor binding at the variant rs2680708 (17q22). Rare variant and PRS associations, with concomitant functional interpretation of risk mechanisms, can help clarify the full genetic architecture of prostate cancer and other complex traits. SIGNIFICANCE: This study maps the biological relationships between diverse risk factors for prostate cancer, integrating different functional datasets to interpret and model genome-wide data from over 200,000 men with and without prostate cancer.See related commentary by Lachance, p. 1637.

Samasy: an automated system for sample selection and robotic transfer.

Sample automation and management is increasingly important as the number and size of population-scale and high-throughput projects grow. This is particularly the case in large-scale population studies where sample size is far outpacing the commonly used 96-well plate format. To facilitate management and transfer of samples in this format, we present Samasy, a web-based application for the construction of a sample database, intuitive display of sample and batch information, and facilitation of automated sample transfer or subset. Samasy is designed with ease-of-use in mind, can be quickly set up, and runs in any web browser.

An open-source method to analyze optokinetic reflex responses in larval zebrafish.

BACKGROUND: Optokinetic reflex (OKR) responses provide a convenient means to evaluate oculomotor, integrative and afferent visual function in larval zebrafish models, which are commonly used to elucidate molecular mechanisms underlying development, disease and repair of the vertebrate nervous system. NEW METHOD: We developed an open-source MATLAB-based solution for automated quantitative analysis of OKR responses in larval zebrafish. The package includes applications to: (i) generate sinusoidally-transformed animated grating patterns suitable for projection onto a cylindrical screen to elicit the OKR; (ii) determine and record the angular orientations of the eyes in each frame of a video recording showing the OKR response; and (iii) analyze angular orientation data from the tracking program to yield a set of parameters that quantify essential elements of the OKR. The method can be employed without modification using the operating manual provided. In addition, annotated source code is included, allowing users to modify or adapt the software for other applications. RESULTS: We validated the algorithms and measured OKR responses in normal larval zebrafish, showing good agreement with published quantitative data, where available. COMPARISON WITH EXISTING METHOD(S): We provide the first open-source method to elicit and analyze the OKR in larval zebrafish. The wide range of parameters that are automatically quantified by our algorithms significantly expands the scope of quantitative analysis previously reported. CONCLUSIONS: Our method for quantifying OKR responses will be useful for numerous applications in neuroscience using the genetically- and chemically-tractable zebrafish model.

Determination of ubiquitin fitness landscapes under different chemical stresses in a classroom setting.

Ubiquitin is essential for eukaryotic life and varies in only 3 amino acid positions between yeast and humans. However, recent deep sequencing studies indicate that ubiquitin is highly tolerant to single mutations. We hypothesized that this tolerance would be reduced by chemically induced physiologic perturbations. To test this hypothesis, a class of first year UCSF graduate students employed deep mutational scanning to determine the fitness landscape of all possible single residue mutations in the presence of five different small molecule perturbations. These perturbations uncover 'shared sensitized positions' localized to areas around the hydrophobic patch and the C-terminus. In addition, we identified perturbation specific effects such as a sensitization of His68 in HU and a tolerance to mutation at Lys63 in DTT. Our data show how chemical stresses can reduce buffering effects in the ubiquitin proteasome system. Finally, this study demonstrates the potential of lab-based interdisciplinary graduate curriculum.

A large multiethnic genome-wide association study of prostate cancer identifies novel risk variants and substantial ethnic differences.

UNLABELLED: A genome-wide association study (GWAS) of prostate cancer in Kaiser Permanente health plan members (7,783 cases, 38,595 controls; 80.3% non-Hispanic white, 4.9% African-American, 7.0% East Asian, and 7.8% Latino) revealed a new independent risk indel rs4646284 at the previously identified locus 6q25.3 that replicated in PEGASUS (N = 7,539) and the Multiethnic Cohort (N = 4,679) with an overall P = 1.0 × 10(-19) (OR, 1.18). Across the 6q25.3 locus, rs4646284 exhibited the strongest association with expression of SLC22A1 (P = 1.3 × 10(-23)) and SLC22A3 (P = 3.2 × 10(-52)). At the known 19q13.33 locus, rs2659124 (P = 1.3 × 10(-13); OR, 1.18) nominally replicated in PEGASUS. A risk score of 105 known risk SNPs was strongly associated with prostate cancer (P < 1.0 × 10(-8)). Comparing the highest to lowest risk score deciles, the OR was 6.22 for non-Hispanic whites, 5.82 for Latinos, 3.77 for African-Americans, and 3.38 for East Asians. In non-Hispanic whites, the 105 risk SNPs explained approximately 7.6% of disease heritability. The entire GWAS array explained approximately 33.4% of heritability, with a 4.3-fold enrichment within DNaseI hypersensitivity sites (P = 0.004). SIGNIFICANCE: Taken together, our findings of independent risk variants, ethnic variation in existing SNP replication, and remaining unexplained heritability have important implications for further clarifying the genetic risk of prostate cancer. Our findings also suggest that there may be much promise in evaluating understudied variation, such as indels and ethnically diverse populations.

Quantification of larval zebrafish motor function in multiwell plates using open-source MATLAB applications.

This article describes a method to quantify the movements of larval zebrafish in multiwell plates, using the open-source MATLAB applications LSRtrack and LSRanalyze. The protocol comprises four stages: generation of high-quality, flatly illuminated video recordings with exposure settings that facilitate object recognition; analysis of the resulting recordings using tools provided in LSRtrack to optimize tracking accuracy and motion detection; analysis of tracking data using LSRanalyze or custom MATLAB scripts; and implementation of validation controls. The method is reliable, automated and flexible, requires <1 h of hands-on work for completion once optimized and shows excellent signal:noise characteristics. The resulting data can be analyzed to determine the following: positional preference; displacement, velocity and acceleration; and duration and frequency of movement events and rest periods. This approach is widely applicable to the analysis of spontaneous or stimulus-evoked zebrafish larval neurobehavioral phenotypes resulting from a broad array of genetic and environmental manipulations, in a multiwell plate format suitable for high-throughput applications.

Identification of fat4 and tsc22d1 as novel candidate genes for spontaneous pulmonary adenomas.

Genetic influences that underlie spontaneous lung oncogenesis are poorly understood. The objective of this study was to determine the genetic influences on spontaneous pulmonary adenoma frequency and severity in 28 strains of mice as part of a large-scale aging study conducted at the Jackson Aging Center (http://agingmice.jax.org/). Genome-wide association studies were conducted in these strains with both low-density (132,000) and high-density (4,000,000) panel of single-nucleotide polymorphisms (SNP). Our analysis revealed that adenomas were relatively less frequent and less severe in females than males, and that loci implicated in frequency and severity were often different between male and female mice. While some of the significant loci identified mapped to genomic locations known to be responsible for carcinogen-induced cancers (e.g., Pas1), others were unique to our study. In particular, Fat4 was influential in males and Tsc22d1 was influential in females. SNPs implicated were predicted to alter amino acid sequence and change protein function. In summary, our results suggested that genetic influences that underlie pulmonary adenoma frequency are dependent on gender, and that Fat4 and Tsc22d1 are likely candidate genes to influence formation of spontaneous pulmonary adenoma in aging male and female mice, respectively.