Interpretation Pipeline Research Articles

Towards interpreting multi-temporal deep learning models in crop mapping

Multi-temporal deep learning approaches have exhibited excellent classification performance in large-scale crop mapping. These approaches efficiently and automatically transform remote sensing time series into high-dimensional feature representations to identify crop types. The lack of interpretation, however, is regarded as a major drawback of these high-performance approaches. Interpreting deep learning approaches in multi-temporal crop mapping is critical for verifying their reliability. This study aims to quantify the impact of multi-temporal information in input time series on classification performance and develop a multi-perspective interpretation pipeline for deep learning models. The pipeline involves three interpretation approaches: evaluating input feature importance, analyzing hidden features, and monitoring temporal changes in model's soft output. An experiment is conducted to classify corn and soybean in the U.S corn belt in 2018. The study area consists of three sites each encompassing millions of pixel-level samples at 30 m resolution. The Landsat Analysis Ready Data are used as the input remote sensing time series and Cropland Data Layer is used as the ground reference. Attention-based Long Short-Term Memory (AtLSTM) and Transformer models are built as multi-temporal deep learning models, and compared to Random Forest (RF). Complete time series input in the correct order achieves a higher overall accuracy of 97.8% than using single-window or out-of-order inputs, indicating multi-temporal information facilitates crop classification. An assessment of the input feature importance demonstrates that the AtLSTM, Transformer, and RF models all consider the period from weeks 11 to 20 (early-July to late-August) as a key growth period and the shortwave infrared band as the critical band for corn and soybean discrimination. Hidden feature analysis suggests that the AtLSTM model accumulates the useful information over the growth period, while the Transformer model extracts the temporal dependencies that contribute important information to high-level feature learning. The learned features contain more effective and refined information than the raw input features and thus are better suited for crop classification. The soft output analysis in the in-season classification scenario demonstrates that increased length of input time series improves the model's confidence in the classification results. The further comparison of input feature importance in different sites and years demonstrates the applicability of the interpretation approach at larger spatiotemporal extents with heterogeneous landscapes and interannual variability. This study provides a multi-perspective evaluation to identify key features in multi-spectral and multi-temporal remote sensing data, and yields a practical approach to integrate agronomy knowledge in deep learning-based crop mapping.

DIVIS: Integrated and Customizable Pipeline for Cancer Genome Sequencing Analysis and Interpretation

Next-generation sequencing (NGS) has drastically enhanced human cancer research, but diverse sequencing strategies, complicated open-source software, and the identification of massive numbers of mutations have limited the clinical application of NGS. Here, we first presented GPyFlow, a lightweight tool that flexibly customizes, executes, and shares workflows. We then introduced DIVIS, a customizable pipeline based on GPyFlow that integrates read preprocessing, alignment, variant detection, and annotation of whole-genome sequencing, whole-exome sequencing, and gene-panel sequencing. By default, DIVIS screens variants from multiple callers and generates a standard variant-detection format list containing caller evidence for each sample, which is compatible with advanced analyses. Lastly, DIVIS generates a statistical report, including command lines, parameters, quality-control indicators, and mutation summary. DIVIS substantially facilitates complex cancer genome sequencing analyses by means of a single powerful and easy-to-use command. The DIVIS code is freely available at https://github.com/niu-lab/DIVIS, and the docker image can be downloaded from https://hub.docker.com/repository/docker/sunshinerain/divis.

A large-scale validation of NOCIt’s a posteriori probability of the number of contributors and its integration into forensic interpretation pipelines

Forensic DNA signal is notoriously challenging to interpret and requires the implementation of computational tools that support its interpretation. While data from high-copy, low-contributor samples result in electropherogram signal that is readily interpreted by probabilistic methods, electropherogram signal from forensic stains is often garnered from low-copy, high-contributor-number samples and is frequently obfuscated by allele sharing, allele drop-out, stutter and noise. Since forensic DNA profiles are too complicated to quantitatively assess by manual methods, continuous, probabilistic frameworks that draw inferences on the Number of Contributors (NOC) and compute the Likelihood Ratio (LR) given the prosecution’s and defense’s hypotheses have been developed.In the current paper, we validate a new version of the NOCIt inference platform that determines an A Posteriori Probability (APP) distribution of the number of contributors given an electropherogram. NOCIt is a continuous inference system that incorporates models of peak height (including degradation and differential degradation), forward and reverse stutter, noise and allelic drop-out while taking into account allele frequencies in a reference population. We established the algorithm’s performance by conducting tests on samples that were representative of types often encountered in practice. In total, we tested NOCIt’s performance on 815 degraded, UV-damaged, inhibited, differentially degraded, or uncompromised DNA mixture samples containing up to 5 contributors. We found that the model makes accurate, repeatable and reliable inferences about the NOCs and significantly outperformed methods that rely on signal filtering.By leveraging recent theoretical results of Slooten and Caliebe (FSI:G, 2018) that, under suitable assumptions, establish the NOC can be treated as a nuisance variable, we demonstrated that when NOCIt’s APP is used in conjunction with a downstream likelihood ratio (LR) inference system that employs the same probabilistic model, a full evaluation across multiple contributor numbers is rendered. This work, therefore, illustrates the power of modern probabilistic systems to report holistic and interpretable weights-of-evidence to the trier-of-fact without assigning a specified number of contributors or filtering signal.

Genetic basis of neurodevelopmental disorders in 103 Jordanian families.

We recruited 103 families from Jordan with neurodevelopmental disorders (NDD) and patterns of inheritance mostly suggestive of autosomal recessive inheritance. In each family, we investigated at least one affected individual using exome sequencing and an in-house diagnostic variant interpretation pipeline including a search for copy number variation. This approach led us to identify the likely molecular defect in established disease genes in 37 families. We could identify 25 pathogenic nonsense and 11 missense variants as well as 3 pathogenic copy number variants and 1 repeat expansion. Notably, 11 of the disease-causal variants occurred de novo. In addition, we prioritized a homozygous frameshift variant in PUS3 in two sisters with intellectual disability. To our knowledge, PUS3 has been postulated only recently as a candidate disease gene for intellectual disability in a single family with three affected siblings. Our findings provide additional evidence to establish loss of PUS3 function as a cause of intellectual disability.

Genome Sequencing Explores Complexity of Chromosomal Abnormalities in Recurrent Miscarriage

Recurrent miscarriage (RM) affects millions of couples globally, and half of them have no demonstrated etiology. Genome sequencing (GS) is an enhanced and novel cytogenetic tool to define the contribution of chromosomal abnormalities in human diseases. In this study we evaluated its utility in RM-affected couples. We performed low-pass GS retrospectively for 1,090 RM-affected couples, all of whom had routine chromosome analysis. A customized sequencing and interpretation pipeline was developed to identify chromosomal rearrangements and deletions/duplications with confirmation by fluorescence in situ hybridization, chromosomal microarray analysis, and PCR studies. Low-pass GS yielded results in 1,077 of 1,090 couples (98.8%) and detected 127 chromosomal abnormalities in 11.7% (126/1,077) of couples; both members of one couple were identified with inversions. Of the 126 couples, 39.7% (50/126) had received former diagnostic results by karyotyping characteristic of normal human male or female karyotypes. Low-pass GS revealed additional chromosomal abnormalities in 50 (4.0%) couples, including eight with balanced translocations and 42 inversions. Follow-up studies of these couples showed a higher miscarriage/fetal-anomaly rate of 5/10 (50%) compared to 21/93 (22.6%) in couples with normal GS, resulting in a relative risk of 2.2 (95% confidence interval, 1.1 to 4.6). In these couples, this protocol significantly increased the diagnostic yield of chromosomal abnormalities per couple (11.7%) in comparison to chromosome analysis (8.0%, chi-square test p=0.000751). In summary, low-pass GS identified underlying chromosomal aberrations in 1 in 9 RM-affected couples, enabling identification of a subgroup of couples with increased risk of subsequent miscarriage who would benefit from a personalized intervention.

Mapping the N-Terminome in Tissue Biopsies by PCT-TAILS.

Proteases play pivotal roles in multiple biological processes in all living organisms and are tightly regulated under normal conditions, but alterations in the proteolytic system and uncontrolled protease activity result in multiple pathological conditions. A disease will most often be defined by an ensemble of cleavage events-a proteolytic signature, thus the system-wide study of protease substrates has gained significant attention and identification of disease specific clusters of protease substrates holds great promise as targets for diagnostics and therapy.In this chapter we describe a method that enables fast and reproducible analysis of protease substrates and proteolytic products in an amount of tissue less than the quantity obtained by a standard biopsy. The method combines tissue disruption and protein extraction by pressure cycling technology (PCT), N-terminal enrichment by tandem mass tag (TMT)-terminal amine isotopic labeling of substrates (TAILS), peptide analysis by mass spectrometry (MS), and a general pipeline for interpretation of the data.

Assessment of blind predictions of the clinical significance of BRCA1 and BRCA2 variants.

Testing for variation in BRCA1 and BRCA2 (commonly referred to as BRCA1/2), has emerged as a standard clinical practice and is helping countless women better understand and manage their heritable risk of breast and ovarian cancer. Yet the increased rate of BRCA1/2 testing has led to an increasing number of Variants of Uncertain Significance (VUS), and the rate of VUS discovery currently outpaces the rate of clinical variant interpretation. Computational prediction is a key component of the variant interpretation pipeline. In the CAGI5 ENIGMA Challenge, six prediction teams submitted predictions on 326 newly-interpreted variants from the ENIGMA Consortium. By evaluating these predictions against the new interpretations, we have gained a number of insights on the state of the art of variant prediction and specific steps to further advance this state of the art.

DNAscan: personal computer compatible NGS analysis, annotation and visualisation

BackgroundNext Generation Sequencing (NGS) is a commonly used technology for studying the genetic basis of biological processes and it underpins the aspirations of precision medicine. However, there are significant challenges when dealing with NGS data. Firstly, a huge number of bioinformatics tools for a wide range of uses exist, therefore it is challenging to design an analysis pipeline. Secondly, NGS analysis is computationally intensive, requiring expensive infrastructure, and many medical and research centres do not have adequate high performance computing facilities and cloud computing is not always an option due to privacy and ownership issues. Finally, the interpretation of the results is not trivial and most available pipelines lack the utilities to favour this crucial step.ResultsWe have therefore developed a fast and efficient bioinformatics pipeline that allows for the analysis of DNA sequencing data, while requiring little computational effort and memory usage. DNAscan can analyse a whole exome sequencing sample in 1 h and a 40x whole genome sequencing sample in 13 h, on a midrange computer. The pipeline can look for single nucleotide variants, small indels, structural variants, repeat expansions and viral genetic material (or any other organism). Its results are annotated using a customisable variety of databases and are available for an on-the-fly visualisation with a local deployment of the gene.iobio platform. DNAscan is implemented in Python. Its code and documentation are available on GitHub: https://github.com/KHP-Informatics/DNAscan. Instructions for an easy and fast deployment with Docker and Singularity are also provided on GitHub.ConclusionsDNAscan is an extremely fast and computationally efficient pipeline for analysis, visualization and interpretation of NGS data. It is designed to provide a powerful and easy-to-use tool for applications in biomedical research and diagnostic medicine, at minimal computational cost. Its comprehensive approach will maximise the potential audience of users, bringing such analyses within the reach of non-specialist laboratories, and those from centres with limited funding available.

Challenges of Providing Concordant Interpretation of Somatic Variants in Non-Small Cell Lung Cancer: A Multicenter Study.

Background: Success of multiple-gene mutation tests by next-generation sequencing (NGS), associated with molecular targeting therapies for cancers, depending on the accuracy and consistency of interpreting variants. Here, we summarized reports from clinical laboratories for cases with non-small cell lung cancer (NSCLC) and discussed conflicting interpretations of somatic variants.Methods: Three mimetic DNA samples, containing six somatic mutations, were prepared based on three clinical case reports of NSCLC. Clinical reports and genetic testing questionnaires were collected from 67 laboratories enrolled in this investigation.Results: Thirty-four laboratories with correct variant results identified two variants, based on FDA approval of targeted drugs for the same tumor, consistently, with strong clinical significance, whereas the other variants were classified with conflicting interpretations. Discordant interpretations were reported for ERBB2 with three different classifications, including strong clinical significance (53.0%, 18/34), potential clinical significance (38.2%, 13/34), and unknown significance (8.8%, 3/34). In the variant therapeutic drug recommendation section, 32.4% of the laboratories (11/34) did not recommend all the available therapeutic drugs designated by the National Comprehensive Cancer Network (NCCN). In the remaining group of 33 laboratories with incorrect variant results, less correct classifications were acquired for the variants with strong clinical significance.Conclusions: Owing to numerous reasons, the interpretation of variants differed greatly, which might in turn lead to the inappropriate clinical care of patients with NSCLC. By analyzing the limitations of different databases used by laboratories, we integrated various types of databases with different levels of evidence to form a comprehensive and detailed variant interpretation pipeline, aiming to standardize the variant classification and provide accurate and sufficient therapeutic drug recommendation to clinicians for minimal-inappropriate therapeutic options.

Fully Automated Echocardiogram Interpretation in Clinical Practice.

Automated cardiac image interpretation has the potential to transform clinical practice in multiple ways, including enabling serial assessment of cardiac function by nonexperts in primary care and rural settings. We hypothesized that advances in computer vision could enable building a fully automated, scalable analysis pipeline for echocardiogram interpretation, including (1) view identification, (2) image segmentation, (3) quantification of structure and function, and (4) disease detection. Using 14 035 echocardiograms spanning a 10-year period, we trained and evaluated convolutional neural network models for multiple tasks, including automated identification of 23 viewpoints and segmentation of cardiac chambers across 5 common views. The segmentation output was used to quantify chamber volumes and left ventricular mass, determine ejection fraction, and facilitate automated determination of longitudinal strain through speckle tracking. Results were evaluated through comparison to manual segmentation and measurements from 8666 echocardiograms obtained during the routine clinical workflow. Finally, we developed models to detect 3 diseases: hypertrophic cardiomyopathy, cardiac amyloid, and pulmonary arterial hypertension. Convolutional neural networks accurately identified views (eg, 96% for parasternal long axis), including flagging partially obscured cardiac chambers, and enabled the segmentation of individual cardiac chambers. The resulting cardiac structure measurements agreed with study report values (eg, median absolute deviations of 15% to 17% of observed values for left ventricular mass, left ventricular diastolic volume, and left atrial volume). In terms of function, we computed automated ejection fraction and longitudinal strain measurements (within 2 cohorts), which agreed with commercial software-derived values (for ejection fraction, median absolute deviation=9.7% of observed, N=6407 studies; for strain, median absolute deviation=7.5%, n=419, and 9.0%, n=110) and demonstrated applicability to serial monitoring of patients with breast cancer for trastuzumab cardiotoxicity. Overall, we found automated measurements to be comparable or superior to manual measurements across 11 internal consistency metrics (eg, the correlation of left atrial and ventricular volumes). Finally, we trained convolutional neural networks to detect hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary arterial hypertension with C statistics of 0.93, 0.87, and 0.85, respectively. Our pipeline lays the groundwork for using automated interpretation to support serial patient tracking and scalable analysis of millions of echocardiograms archived within healthcare systems.

Validation of the Oncomine\u2122 focus panel for next-generation sequencing of clinical tumour samples

The clinical utility of next-generation sequencing (NGS) for a diverse range of targets is expanding, increasing the need for multiplexed analysis of both DNA and RNA. However, translation into daily use requires a rigorous and comprehensive validation strategy. The aim of this clinical validation was to assess the performance of the Ion Torrent Personal Genome Machine (IonPGM™) and validate the Oncomine™ Focus DNA and RNA Fusion panels for clinical application in solid tumour testing of formalin-fixed, paraffin-embedded (FFPE) tissue. Using a mixture of routine FFPE and reference material across a variety of tissue and specimen types, we sequenced 86 and 31 samples on the Oncomine™ Focus DNA and RNA Fusion assays, respectively. This validation considered a number of parameters including the clinical robustness of the bioinformatics pipeline for variant detection and interpretation. The Oncomine™ Focus DNA assay had a sample and variant-based sensitivity of 99.1 and 97.1%, respectively, and an assay specificity of 100%. The Oncomine™ Focus Fusion panel had a good sensitivity and specificity based upon the samples assessed, however requires further validation to confirm findings due to limited sample numbers. We observed a good sequencing performance based upon amplicon, gene (hotspot variants within gene) and sample specific analysis with 92% of clinical samples obtaining an average amplicon coverage above 500X. Detection of some indels was challenging for the routine IonReporter™ workflow; however, the addition of NextGENe® software improved indel identification demonstrating the importance of both bench and bioinformatic validation. With an increasing number of clinically actionable targets requiring a variety of methodologies, NGS provides a cost-effective and time-saving methodology to assess multiple targets across different modalities. We suggest the use of multiple analysis software to ensure identification of clinically applicable variants.

Abstract 880: A computational and statistical approach for interpreting real-time in-vitro gene reporter data

Abstract We have developed a new computational approach to track and analyze data from cells engineered with fluorescent cell cycle gene reporters. Gene reporters are a rapidly growing tool to gather information captured with time-lapse imaging, and can give insights to real-time metabolic conditions, cell cycle state, cancer stem cell state, and many other applications. Gene reporters may reflect cellular conditions that are cyclical rather than monotonic. Currently, the primary mechanism for interpretation of time lapse gene reporter data is reporting the mean fluorescence of images over time. This simplistic analysis discards information such as the cycle length and variation between of gene reporters between cells, and can be biased due to the overall fluorescence shift from population growth. More nuanced data acquisition and analysis approaches have been limited due to requirement of laborious hand-on data analysis. We have implemented a high-throughput computational pipeline for interpretation of cell cycle information using a combination of commercial and newly developed software to partition the activities of individual cells. The fluorescence information for each cell is smoothed and the cycle states for each cell are assigned adaptively. Our approach captures and quantifies the total time each individual cell spends in each presumptive state, and then analyses state information from the cohort of observed cells using censored survival time data methods such as the Cox Proportional Hazards Model to account for cell death or movement out of the observed area. As a proof of concept for our pipeline, we engineered 3T3 fibroblast cells with a cell cycle reporter construct including mCherry-CDT1 and mVenous-p27K- reporters which can discriminate cells in G0 phase from G1 phase. We calculated the sample size required for statistical power of 80% to detect a hazard ratio (HR) of 0.8 between groups at α = 0.05 at 386 tracked cells per group. As our treatment model, we used the Protein Phosphatase 2A inhibitor Okadaic Acid (OA) an agent known to promote departure from G0. We captured a distinct total of 601 vehicle cells and 479 treatment cells in a 24 hour monitoring period with 20 minute capture intervals. We were able to discern subtle differences between control and treatment G0 time, including a subtle, statistically significant decrease in G0 time after treatment with OA (p=0.0166, HR = 0.84). Here we have developed a computational pipeline with the capability for high-throughput image-stack data processing and nuanced quantification. Our approach assigns cycling cell states adaptively and the time until each cell departs its current state is tallied and interpreted statistically with survival time statistical analysis. The cell-level analysis affords a high degree of statistical power to discriminate group differences. Our pipeline is highly adaptable and has a multitude of applications in cancer biology research. Citation Format: Alexander T. Pearson, Dan Sun, Zhengda Li, Jacques Nor, Qiong Yang, Laura Buttitta. A computational and statistical approach for interpreting real-time in-vitro gene reporter data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 880. doi:10.1158/1538-7445.AM2017-880

Multi-purpose ECG telemetry system

BackgroundThe Electrocardiogram ECG is one of the most important non-invasive tools for cardiac diseases diagnosis. Taking advantage of the developed telecommunication infrastructure, several approaches that address the development of telemetry cardiac devices were introduced recently. Telemetry ECG devices allow easy and fast ECG monitoring of patients with suspected cardiac issues. Choosing the right device with the desired working mode, signal quality, and the device cost are still the main obstacles to massive usage of these devices.MethodsIn this paper, we introduce design, implementation, and validation of a multi-purpose telemetry system for recording, transmission, and interpretation of ECG signals in different recording modes. The system consists of an ECG device, a cloud-based analysis pipeline, and accompanied mobile applications for physicians and patients. The proposed ECG device’s mechanical design allows laypersons to easily record post-event short-term ECG signals, using dry electrodes without any preparation. Moreover, patients can use the device to record long-term signals in loop and holter modes, using wet electrodes. In order to overcome the problem of signal quality fluctuation due to using different electrodes types and different placements on subject’s chest, customized ECG signal processing and interpretation pipeline is presented for each working mode.ResultsWe present the evaluation of the novel short-term recorder design. Recording of an ECG signal was performed for 391 patients using a standard 12-leads golden standard ECG and the proposed patient-activated short-term post-event recorder. In the validation phase, a sample of validation signals followed peer review process wherein two experts annotated the signals in terms of signal acceptability for diagnosis.We found that 96% of signals allow detecting arrhythmia and other signal’s abnormal changes. Additionally, we compared and presented the correlation coefficient and the automatic QRS delineation results of both short-term post-event recorder and 12-leads golden standard ECG recorder.ConclusionsThe proposed multi-purpose ECG device allows physicians to choose the working mode of the same device according to the patient status. The proposed device was designed to allow patients to manage the technical requirements of both working modes. Post-event short-term ECG recording using the proposed design provide physicians reliable three ECG leads with direct symptom-rhythm correlation.

Mesh Is More—Using All Geometric Dimensions for the Archaeological Analysis and Interpretative Mapping of 3D Surfaces

In various disciplines, particularly those that utilise techniques developed in the geosciences, the display, analysis and interpretation of three-dimensional (3D) data is very important. This is also the case in archaeology. Irrespective of a site- or landscape-centred point of focus, archaeology deals with very complex surfaces and always examines traces of past human presence in three geometrical dimensions. Visualising these detailed geometric environments is not that much of an issue anymore; however, interactively interpreting and mapping them is still problematic. Despite the steady increase in technologies to create 3D models, there is still a serious lack of tools that allow for easy interaction with these models in a metrical and coordinate system-aware environment. As a result, most—if not all—interpretation workflows will first downscale 3D data to two-and-a-half-dimensional (2.5D) or 2D data sets, thus effectively discarding up to one geometrical dimension. To enable or enhance the perception of topographic characteristics in these geometrically compromised datasets, various visualisation techniques have been developed to artificially restore and enhance the data that was initially discarded. While these techniques work very well to enhance the remaining pertinent features present in such data sets, data downscaling can nevertheless irrevocably eliminate significant amounts of important archaeological information. Therefore, this paper outlines a new processing and interpretation pipeline for complex archaeological 3D surfaces that do not rely on downscaling of data, while also discussing several 3D-related concepts and issues along the way. More specifically, this article focusses on the generation of intelligently decimated, two-manifold triangular meshes and the subsequent geo-referenced 3D interpretative mapping of these surfaces. Furthermore, all applications can be considered low- and even no-cost, making this a readily implemented processing and interpretation workflow. Additionally, all software packages are easy to learn and flexible enough for implementation in any existing mixed software pipeline.

Genetics of Combined Pituitary Hormone Deficiency: Roadmap into the Genome Era.

The genetic basis for combined pituitary hormone deficiency (CPHD) is complex, involving 30 genes in a variety of syndromic and nonsyndromic presentations. Molecular diagnosis of this disorder is valuable for predicting disease progression, avoiding unnecessary surgery, and family planning. We expect that the application of high throughput sequencing will uncover additional contributing genes and eventually become a valuable tool for molecular diagnosis. For example, in the last 3 years, six new genes have been implicated in CPHD using whole-exome sequencing. In this review, we present a historical perspective on gene discovery for CPHD and predict approaches that may facilitate future gene identification projects conducted by clinicians and basic scientists. Guidelines for systematic reporting of genetic variants and assigning causality are emerging. We apply these guidelines retrospectively to reports of the genetic basis of CPHD and summarize modes of inheritance and penetrance for each of the known genes. In recent years, there have been great improvements in databases of genetic information for diverse populations. Some issues remain that make molecular diagnosis challenging in some cases. These include the inherent genetic complexity of this disorder, technical challenges like uneven coverage, differing results from variant calling and interpretation pipelines, the number of tolerated genetic alterations, and imperfect methods for predicting pathogenicity. We discuss approaches for future research in the genetics of CPHD.

Accurate genetic diagnosis of Finnish pulmonary arterial hypertension patients using oligonucleotide-selective sequencing.

The genetic basis of pulmonary arterial hypertension (PAH) among Finnish PAH patients is poorly understood. We adopted a novel-targeted next-generation sequencing (NGS) approach called Oligonucleotide-Selective Sequencing (OS-Seq) and developed a custom data analysis and interpretation pipeline to identify pathogenic base substitutions, insertions, and deletions in seven genes associated with PAH (BMPR2, BMPR1B, ACVRL1, ENG, SMAD9, CAV1, and KCNK3) from Finnish PAH patients. This study represents the first clinical study with OS-Seq technology on patients suffering from a rare genetic disorder. We analyzed DNA samples from 21 Finnish PAH patients, whose BMPR2 and ACVRL1 mutation status had been previously studied using Sanger sequencing. Our sequencing panel covered 100% of the targeted base pairs with >15× sequencing depth. Pathogenic base substitutions were identified in the BMPR2 gene in 29% of the Finnish PAH cases. Two of the pathogenic variant-positive patients had been previously tested negative using Sanger sequencing. No clinically significant variants were identified in the six other PAH genes. Our study validates the use of targeted OS-Seq for genetic diagnostics of PAH and revealed pathogenic variants that had been previously missed using Sanger sequencing.

Abstract 4260: Standardized decision support in next-generation sequencing (NGS) reports of somatic cancer variants

Abstract Introduction: Of hundreds to thousands of somatic mutations that exist in each cancer genome, a large number are unique and non-recurrent variants. Many such variants occur in tumor genes that have well-established biological and clinical relevance and are putative targets of molecular therapy, however, most variants are still of unknown significance. Prioritizing and reporting genetic variants identified via NGS technologies remains a major challenge. Methods: Based on a systematic framework for cancer variant annotation/prioritization, we propose a structured molecular pathology report using standardized terminology in order to best inform oncology clinical practice. In parallel with the variant interpretation pipeline, we developed a comprehensive knowledge database (Kdb) that integrates tumor types, genes, variants, response/resistance patterns to approved and experimental agents and PubMed identifiers. Results: As physicians rely heavily on limited subsets of data and frequently apply rules based on strength of evidence in the decision-making process, NGS results are presented in a very discrete manner, avoiding detailed descriptions of each genomic aberration. Reportable variants are grouped in three categories: (i) actionable, which support treatment recommendation, enrollment in clinical trials and/or have prognostic or diagnostic implications; (ii) biologically relevant but not clearly actionable, including novel variants lacking functional preclinical validation in known cancer genes; and (iii) variants of unknown significance. Predictive associations are reported according to tumor type and classified in a hierarchical way based on the strength of evidence: (i) late trials; (ii) early trials; (iii) case reports; and (iv) preclinical data being explored in clinical trials. In our internal Kdb, tumor types with the largest number of emerging predictive associations are acute leukemias, non-small cell lung cancer, brain tumors, melanoma, colorectal, ovarian and breast cancers, comprising more than 100 unique gene - drug interactions. Further interpretation of gene variants requires a team with strong background in cancer biology, careful evaluation of the published literature and ability to match a patient's tumor genotype to clinical trials in the context of multiple aberrations. Conclusions: With large amounts of data being generated as high throughput sequencing assays enter the clinical realm, there is a growing need to better communicate relevant findings in a timely manner while remaining cognizant of the potential consequences of misuse or overinterpretation of genomic information. The scientific knowledge on actionable events should be presented the report, so that physicians can make evidence-based decisions and patients may benefit from treatment with matched targeted agents. We hope that our experience in this process will help other institutions implement similar programs. Citation Format: Rodrigo Dienstmann, Fei Dong, Darrell Borger, Dora Dias Santagata, Leif W. Ellisen, Long P. Le, A. John Iafrate. Standardized decision support in next-generation sequencing (NGS) reports of somatic cancer variants. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4260. doi:10.1158/1538-7445.AM2014-4260

The Interlaboratory Robustness Of Next-Generation Sequencing (IRON) Study Phase II: Deep-Sequencing Analyses Of Hematological Malignancies Performed In 8,867 Cases By An International Network Involving 27 Laboratories

IntroductionAmplicon deep-sequencing using second-generation sequencing technology is an innovative molecular diagnostic technique and enables a highly-sensitive detection of mutations. As an international consortium we had investigated previously the robustness, precision, and reproducibility of 454 amplicon next-generation sequencing (NGS) across 10 laboratories from 8 countries (Leukemia, 2011;25:1840-8). AimsIn Phase II of the study, we established distinct working groups for various hematological malignancies, i.e. acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CLL), chronic myelogenous leukemia (CML), myelodysplastic syndromes (MDS), myeloproliferative neoplasms (MPN), and multiple myeloma. Currently, 27 laboratories from 13 countries are part of this research consortium. In total, 74 gene targets were selected by the working groups and amplicons were developed for a NGS deep-sequencing assay (454 Life Sciences, Branford, CT). A data analysis pipeline was developed to standardize mutation interpretation both for accessing raw data (Roche Amplicon Variant Analyzer, 454 Life Sciences) and variant interpretation (Sequence Pilot, JSI Medical Systems, Kippenheim, Germany). ResultsWe will report on the design, standardization, quality control aspects, landscape of mutations, as well as the prognostic and predictive utility of this assay in a cohort of 8,867 cases. Overall, 1,146 primer sequences were designed and tested. In detail, for example in AML, 924 cases had been screened for CEBPA mutations. RUNX1 mutations were analyzed in 1,888 cases applying the deep-sequencing read counts to study the stability of such mutations at relapse and their utility as a biomarker to detect residual disease. Analyses of DNMT3A (n=1,041) were focused to perform landscape investigations and to address the prognostic relevance. Additionally, this working group is focusing on TET2, ASXL1, and TP53 analyses. A novel prognostic model is being developed allowing stratification of AML into prognostic subgroups based on molecular markers only. In ALL, 1,124 pediatric and adult cases have been screened, including 763 assays for TP53 mutations both at diagnosis and relapse of ALL. Pediatric and adult leukemia expert labs developed additional content to study the mutation incidence of other B and T lineage markers such as IKZF1, JAK2, IL7R, PAX5, EP300, LEF1, CRLF2, PHF6, WT1, JAK1, PTEN, AKT1, IL7R, NOTCH1, CREBBP, or FBXW7. Further, the molecular landscape of CLL is changing rapidly. As such, a separate working group focused on analyses including NOTCH1, SF3B1, MYD88, XPO1, FBXW7 and BIRC3. Currently, 922 cases were screened to investigate the range of mutational burden of NOTCH1 mutations for their prognostic relevance. In MDS, RUNX1 mutation analyses were performed in 977 cases. The prognostic relevance of TP53 mutations in MDS was assessed in additional 327 cases, including isolated deletions of chromosome 5q. Next, content was developed targeting genes of the cellular splicing component, e.g. SF3B1, SRSF2, U2AF1, and ZRSR2. In BCR-ABL1-negative MPN, nine genes of interest (JAK2, MPL, TET2, CBL, KRAS, EZH2, IDH1, IDH2, ASXL1) have been analyzed in a cohort of 155 primary myelofibrosis cases searching for novel somatic mutations and addressing their relevance for disease progression and leukemia transformation. Moreover, an assay was developed and applied to CMML cases allowing the simultaneous analysis of 25 leukemia-associated target genes in a single sequencing run using just 20 ng of starting DNA. Finally, nine laboratories are studying CML, applying ultra-deep sequencing of the BCR-ABL1 tyrosine kinase domain. Analyses were performed on 615 cases investigating the dynamics of expansion of mutated clones under various tyrosine kinase inhibitor therapies. ConclusionMolecular characterization of hematological malignancies today requires high diagnostic sensitivity and specificity. As part of the IRON-II study, a network of laboratories analyzed a variety of disease entities applying amplicon-based NGS assays. Importantly, the consortium not only standardized assay design for disease-specific panels, but also achieved consensus on a common data analysis pipeline for mutation interpretation. Distinct working groups have been forged to address scientific tasks and in total 8,867 cases had been analyzed thus far. Disclosures:Kohlmann:Roche Diagnostics: Honoraria; MLL Munich Leukemia Laboratory: Employment. Martinelli:Roche Diagnostics: Research Support Other. Alikian:Roche Diagnostics: Research Support Other. Auber:Roche Diagnostics: Research Support Other. Belickova:Roche Diagnostics: Research Support Other. Bronzini:Roche Diagnostics: Research Support Other. Cazzaniga:Roche Diagnostics: Research Support Other. Chiaretti:Roche Diagnostics: Research Support Other. Ernst:Roche Diagnostics: Research Support Other. Fuellgrabe:Roche Diagnostics: Research Support Other. Gabriel:Roche Diagnostics: Research Support Other. Hernandez:Roche Diagnostics: Research Support Other. Jansen:Roche Diagnostics: Research Support Other. Iacobucci:Roche Diagnostics: Research Support Other. Lion:Roche Diagnostics: Research Support Other. Lode:Roche Diagnostics: Research Support Other. Martinez-Lopez:Roche Diagnostics: Research Support Other. Mills:Roche Diagnostics: Research Support Other. Mossner:Roche Diagnostics: Research Support Other. Machova Polakova:Roche Diagnostics: Research Support Other. Porret:Roche Diagnostics: Research Support Other. Pospisilova:Roche Diagnostics: Research Support Other. Preudhomme:Roche Diagnostics: Research Support Other. Sayitoglu:Roche Diagnostics: Research Support Other. Soverini:Roche Diagnostics: Research Support Other. Spinelli:Roche Diagnostics: Research Support Other. Thiede:Roche Diagnostics: Research Support Other. Vandenberghe:Roche Diagnostics: Research Support Other. Yeoh:Roche Diagnostics: Research Support Other. Hochhaus:Roche Diagnostics: Research Support Other. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership; Roche Diagnostics: Research Funding.

Algorithms and Processing Pipeline For Error Correction and Detection Of Significant Mutations In The Kinase Domain Of BCR-ABL Analyzed By Next-Generation Sequencing: Implications For Clinical Practice Of Chronic Myeloid Leukemia

High sequencing depths of NGS may cause false-positive variant calls of minor subclones (up to 10%). Errors inserted into the NGS pipeline during sample preparation and sequencing manifest by erroneous detections of mutations including point mutations. Thus, there is a need for algorithms for filtering data which occur by error from relevant biological information.We developed algorithms and a processing pipeline for error correction and detection of significant mutations after NGS of BCR-ABL kinase domain (KD). We validated our algorithms on the retrospective NGS analysis of 135 samples from 15 CML patients in chronic phase (median 8 samples per patient; range 5-19), who developed resistant mutations (confirmed by Sanger Sequencing, SS) during 2-4 lines of therapy.Amplicon libraries were prepared using reverse transcription and 2-step PCR. The second PCR was performed partly using fusion-primers designed within the IRON-II study research consortium (Roche Applied Science) which tested 4 overlapping amplicons and partly using alternative in-house set of fusion-primers that we have developed upfront and which utilized 3 overlapping amplicons covering the KD coding region. Key concept of our error control algorithm was to apply statistics used for bacterial mutation rate prediction, Lea-Coulson probability distribution (Lea and Coulson, J of Genet 1949), to distinguish sequencing pipeline errors from biologically relevant mutations. We postulated that spontaneous mutations in bacteria are similar phenomena as enzyme errors in vitro. In both processes there are new generations of bacteria or transcripts in which mutations or errors replicate exponentially. The error rate distributions based on analysis of c-ABL kinase domain of healthy donors (n=24) were fitted to Lea-Coulson distribution. From this analysis we derived, for each type of single nucleotide substitution, estimated thresholds based on which a particular mutation may be called significant by a self-developed statistical test. We cross-checked our results with results of standard Roche pipeline including GS Amplicon Variant Analyzer.Table 1 summarizes the estimated thresholds to be applied for transitions and transversions. Higher frequency of errors was found in case of using a 3-amplicon assay in comparison to a 4-amplicon assay. The PCR products in the 3-amplicon assay are 71 bp longer on average than in the 4-amplicon assay, thus the error frequency distribution may be dependent on the length of the sequence amplified. Using our algorithm we processed NGS data and reported significant mutations. Overall, no significant mutation that caused resistance during the treatment was detected at the time of diagnosis. During 1st line imatinib treatment 10 resistant mutations in 9 patients were detected as significant 2-5 months earlier than by SS. At the time of therapy switchover, in 3 patients the algorithm already detected minor populations of one of significant mutations F317L, T315I and M351T, while SS did not. These mutations manifested after the therapy switchover and caused treatment failure. After the therapy switch, baseline mutations were still significantly detectable by our algorithm in NGS data, but not by SS in 7 patients who achieved at the time of the analysis PCgR and MMR. In 5 patients, who subsequently failed therapy after switchover, resistant mutations were significantly detected by our algorithm in NGS data 2-9 months earlier than by SS. New minor mutations were revealed by NGS after the therapy switch in 8 patients.Table 1TRANSITIONS (%)TRANSVERSIONS (%)P valueA/GG/AT/CC/TA/C+C/A+T/A+A/T+T/G+G/T+C/G+G/C3-amplicon assay0.0112.24.5311.84.770.570.053.031.032.931.100.134-amplicon assay0.015.171.934.501.930.130.051.200.431.030.430.03Since enzymes create errors during reverse transcription, 2-step PCR and sequencing process, the error correction is an essential part of the bioinformatics pipeline for relevant interpretation of BCR-ABL KD mutations detected with the highly sensitive NGS assay. Our validated algorithm and processing pipeline for significant mutation evaluations from NGS data is helpful for future clinical practice as it filters errors and allows reporting only significant mutations. This avoids false-positive results and misleading interpretations which may negatively influence treatment management of CML patients.Supported by IGANT11555 and NT13899 Disclosures:Machova Polakova:Bristol-Myers Squibb: Honoraria, Research Funding; Novartis: Honoraria, Research Funding. Soverini:Novartis: Consultancy; Bristol-Myers Squibb: Consultancy; ARIAD: Consultancy. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Martinelli:NOVARTIS, BMS(Consultancy and speaker bureau), PFIZER, ARIAD ( Consultancy): Consultancy, Speakers Bureau. Kohlmann:MLL Munich Leukemia Laboratory: Employment; Roche Diagnostics: Honoraria. Klamova:Novartis: Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding.

New Frontiers for Organismal Biology

Understanding how complex organisms function as integrated units that constantly interact with their environment is a long-standing challenge in biology. To address this challenge, organismal biology reveals general organizing principles of physiological systems and behavior—in particular, in complex multicellular animals. Organismal biology also focuses on the role of individual variability in the evolutionary maintenance of diversity. To broadly advance these frontiers, cross-compatibility of experimental designs, methodological approaches, and data interpretation pipelines represents a key prerequisite. It is now possible to rapidly and systematically analyze complete genomes to elucidate genetic variation associated with traits and conditions that define individuals, populations, and species. However, genetic variation alone does not explain the varied individual physiology and behavior of complex organisms. We propose that such emergent properties of complex organisms can best be explained through a renewed emphasis on the context and life-history dependence of individual phenotypes to complement genetic data.