True Variants Research Articles

Improving the Utility of a Cancer Gene Mutation Panel for Multiple Myeloma Patients

Abstract Multiple myeloma is a plasma cell malignancy that accounts for 1-2% of all cancer diagnoses and 10-15% of hematopoietic neoplasms. Clonal rearrangement of the IG locus suggests a monoclonal model for pathogenesis. The presence of cytogenetic abnormalities in >90% of cases suggests transformation by chromosomal instability and rearrangement. Classification and consensus guidelines by The International Myeloma Working Group (IMWG) and others reflect this by categorizing myeloma based on ploidy, translocations, and chromosomal abnormalities. Interestingly, a portion of Emory’s test volume for cancer gene mutation profiling by second generation short-read sequencing is for multiple myeloma cases. This suggests our 75-gene myeloid neoplasm panel (MMP75) is being utilized for medical purposes that are not met by cytogenetic and microarray characterizations performed during the diagnostic workup of multiple myeloma. Inspired by this observation, this study investigates providers motivations for ordering MMP75 and assesses the adequacy of our current implementation for their stated purposes. Anecdotes from ordering physicians indicate interest in MMP75 results when informing prognosis, selecting treatments, and evaluating for treatment associated secondary neoplasms. A more formal survey is underway that will statistically capture how the results of MMP75 testing were ultimately used: to inform diagnosis, determine prognosis, guide treatment, explain progression, identify/rule-out secondary neoplasia, or no effect on management. We also explore the variants called by MMP75 when performed using bone marrow from patients who received a cytogenetic workup for multiple myeloma at some point. A retrospective analysis of the 5-year interval between Jan 2019 and Jan 2024 identified 156 cases meeting this criterion. The number of pathogenic variants per case ranged between 1 and 8, with mutations detected across 42 genes. The genes most frequently identified with pathogenic mutations were TP53 (22.4%), DNMT3A (19.2%), TET2 (14.7%), and ASXL1 (12.2%). All four have been associated with adverse outcomes in patients with multiple myeloma. Activating events in therapeutic targets included mutations in the RAS/RAF pathway (16.7%), JAK/STAT signaling (2.6%), and the kinase KIT (0.6%). While these findings support the use of MMP75 for detection of relevant mutations, the genetic variants identified are not unique to multiple myeloma. Mutations detected in DNMT3A, TET2, and ASXL1 are common in clonal hematopoiesis of indeterminant potential (CHIP) and myeloid neoplasms. Hotspot mutations in TP53 and activating events in the RAS/RAF pathway also occur in various hematologic malignancies. This test therefore detects events that occur in both the plasma cell compartment and other cells of origin. This encumbers distinguishing variants that inform on a patient’s multiple myeloma from events that suggest a secondary neoplasm. Future directions include determining whether the effect of plasma cell enrichment on a mutations variant allele frequency (VAF) in repeat testing can help distinguish true multiple myeloma variants from events in other cells.

Cancer associated variant enrichment CAVE, a gene agnostic approach to identify low burden variants in chronic lymphocytic leukemia

Intratumoral heterogeneity is an important clinical challenge because low burden clones expressing specific genetic alterations drive therapeutic resistance mechanisms. We have developed CAVE (cancer-associated variant enrichment), a gene-agnostic computational tool to identify specific enrichment of low-burden cancer driver variants in next-generation sequencing (NGS) data. For this study, CAVE was applied to TP53 in chronic lymphocytic leukemia (CLL) as a cancer model. Indeed, as TP53 mutations are part of treatment decision-making algorithms and low-burden variants are frequent, there is a need to distinguish true variants from background noise. Recommendations have been published for reliable calling of low-VAF variants of TP53 in CLL and the assessment of the background noise for each platform is essential for the quality of the testing. CAVE is able to detect specific enrichment of low-burden variants starting at variant allele frequencies (VAFs) as low as 0.3%. In silico TP53 dependent and independent analyses confirmed the true driver nature of all these variants. Orthogonal validation using either ddPCR or NGS analyses of follow-up samples confirmed variant identification. CAVE can be easily deployed in any cancer-related NGS workflow to detect the enrichment of low-burden variants of clinical interest.

Investigation of cryptic JAG1 splice variants as a cause of Alagille syndrome and performance evaluation of splice predictor tools

Haploinsufficiency of JAG1 is the primary cause of Alagille syndrome (ALGS), a rare, multisystem disorder. The identification of JAG1 intronic variants outside of the canonical splice region as well as missense variants, both of which lead to uncertain associations with disease, confuses diagnostics. Strategies to determine whether these variants affect splicing include the study of patient RNA or minigene constructs, which are not always available or can be laborious to design, as well as the utilization of computational splice prediction tools. These tools, including SpliceAI and Pangolin, use algorithms to calculate the probability that a variant results in a splice alteration, expressed as a Δ score, with higher Δ scores (>0.2 on a 0-1 scale) positively correlated with aberrant splicing. We studied the consequence of 10 putative splice variants in ALGS patient samples through RNA analysis and compared this to SpliceAI and Pangolin predictions. We identified eight variants with aberrant splicing, seven of which had not been previously validated. Combining these data with non-canonical and missense splice variants reported in the literature, we identified a predictive threshold for SpliceAI and Pangolin with high sensitivity (Δ score >0.6). Moreover, we showed reduced specificity for variants with low Δ scores (<0.2), highlighting a limitation of these tools that results in the misidentification of true splice variants. These results improve genomic diagnostics for ALGS by confirming splice effects for seven variants and suggest that the integration of splice prediction tools with RNA analysis is important to ensure accurate clinical variant classifications.

A Conceding Position for a True Heterozygote Variant Misjudge as Tri-Allele for Adjacent Genetic Marker: “Neighboring” Leads to Misinterpretation

A Conceding Position for a True Heterozygote Variant Misjudge as Tri-Allele for Adjacent Genetic Marker: “Neighboring” Leads to Misinterpretation

Finemap-MiXeR: A variational Bayesian approach for genetic finemapping.

Genome-wide association studies (GWAS) implicate broad genomic loci containing clusters of highly correlated genetic variants. Finemapping techniques can select and prioritize variants within each GWAS locus which are more likely to have a functional influence on the trait. Here, we present a novel method, Finemap-MiXeR, for finemapping causal variants from GWAS summary statistics, controlling for correlation among variants due to linkage disequilibrium. Our method is based on a variational Bayesian approach and direct optimization of the Evidence Lower Bound (ELBO) of the likelihood function derived from the MiXeR model. After obtaining the analytical expression for ELBO's gradient, we apply Adaptive Moment Estimation (ADAM) algorithm for optimization, allowing us to obtain the posterior causal probability of each variant. Using these posterior causal probabilities, we validated Finemap-MiXeR across a wide range of scenarios using both synthetic data, and real data on height from the UK Biobank. Comparison of Finemap-MiXeR with two existing methods, FINEMAP and SuSiE RSS, demonstrated similar or improved accuracy. Furthermore, our method is computationally efficient in several aspects. For example, unlike many other methods in the literature, its computational complexity does not increase with the number of true causal variants in a locus and it does not require any matrix inversion operation. The mathematical framework of Finemap-MiXeR is flexible and may also be applied to other problems including cross-trait and cross-ancestry finemapping.

Detection of a major Lynch Syndrome-causing MLH1 founder variant in a large-scale genotyped cohort

Some 50% of Finnish Lynch Syndrome (LS) cases are caused by a founder variant in MLH1, in which the entire exon 16 has been lost due to an Alu-mediated recombination event. We piloted detecting the variant in FinnGen, a large genotyped cohort comprising approximately 10% of the current Finnish population, and validated the MLH1 founder variant status of identified individuals residing in the Central Finland Biobank catchment area. A consensus sequence flanking the deletion was identified in whole genome sequences of six LS individuals with the founder variant. Genotype data of 212,196 individuals was queried for regional matches to the consensus sequence. Enrichment of cancer and age at cancer onset was compared between matching and non-matching individuals. Variant status was validated for a subset of the identified individuals using a polymerase chain reaction assay. Allelic matches in a chosen target region was detected in 348 individuals, with 89 having a cancer diagnosis (Bonferroni-adjusted p-value = 1), 20 a familial cancer history (p-adj. < .001), with mean age of onset of cancer being 53.6 years (p-adj. = .002). Eighteen of potential variant carriers had been sampled by the Central Finland Biobank, of which four (22%) were validated as true variant carriers. The workflow we have employed identifies MLH1 exon 16 deletion variant carriers from population-wide SNP genotyping data. An alternative design will be sought to limit false positive findings. Large genotyped cohorts provide a potential resource for identification and prevention of hereditary cancer.

Negligible effects of read trimming on the accuracy of germline short variant calling in the human genome

Background Next generation sequencing (NGS) has become a standard tool in the molecular diagnostics of Mendelian disease, and the precision of such diagnostics is greatly affected by the accuracy of variant calling from sequencing data. Recently, we have comprehensively evaluated the performance of multiple variant calling pipelines. However, no systematic analysis of the effects of read trimming on variant discovery with modern variant calling software has yet been performed. Methods In this work, we systematically evaluated the effects of adapters on the performance of 8 variant calling and filtering methods using 14 standard reference Genome-in-a-Bottle (GIAB) samples. Variant calls were compared to the ground truth variant sets, and the effect of adapter trimming with different tools was assessed using major performance metrics (precision, recall, and F1 score). Results We show that adapter trimming has no effect on the accuracy of the best-performing variant callers (e.g., DeepVariant) on whole-genome sequencing (WGS) data. For whole-exome sequencing (WES) datasets subtle improvement of accuracy was observed in some of the samples. In high-coverage WES data (~200x mean coverage), adapter removal allowed for discovery of 2-4 additional true positive variants in only two out of seven datasets tested. Moreover, this effect was not dependent on the median insert size and proportion of adapter sequences in reads. Surprisingly, the effect of trimming on variant calling was reversed when moderate coverage (~80-100x) WES data was used. Finally, we show that some of the recently developed machine learning-based variant callers demonstrate greater dependence on the presence of adapters in reads. Conclusions Taken together, our results indicate that adapter removal is unnecessary when calling germline variants, but suggest that preprocessing methods should be carefully chosen when developing and using machine learning-based variant analysis methods.

Hierarchical joint analysis of marginal summary statistics-Part I: Multipopulation fine mapping and credible set construction.

Recent advancement in genome-wide association studies (GWAS) comes from not only increasingly larger sample sizes but also the shift in focus towards underrepresented populations. Multipopulation GWAS increase power to detect novel risk variants and improve fine-mapping resolution by leveraging evidence and differences in linkage disequilibrium (LD) from diverse populations. Here, we expand upon our previous approach for single-population fine-mapping through Joint Analysis of Marginal SNP Effects (JAM) to a multipopulation analysis (mJAM). Under the assumption that true causal variants are common across studies, we implement a hierarchical model framework that conditions on multiple SNPs while explicitly incorporating the different LD structures across populations. The mJAM framework can be used to first select index variants using the mJAM likelihood with different feature selection approaches. In addition, we present a novel approach leveraging the ideas of mediation to construct credible sets for these index variants. Construction of such credible sets can be performed given any existing index variants. We illustrate the implementation of the mJAM likelihood through two implementations: mJAM-SuSiE (a Bayesian approach) and mJAM-Forward selection. Through simulation studies based on realistic effect sizes and levels of LD, we demonstrated that mJAM performs well for constructing concise credible sets that include the underlying causal variants. In real data examples taken from the most recent multipopulation prostate cancer GWAS, we showed several practical advantages of mJAM over other existing multipopulation methods.

An AI-based approach driven by genotypes and phenotypes to uplift the diagnostic yield of genetic diseases.

Identifying disease-causing variants in Rare Disease patients' genome is a challenging problem. To accomplish this task, we describe a machine learning framework, that we called "Suggested Diagnosis", whose aim is to prioritize genetic variants in an exome/genome based on the probability of being disease-causing. To do so, our method leverages standard guidelines for germline variant interpretation as defined by the American College of Human Genomics (ACMG) and the Association for Molecular Pathology (AMP), inheritance information, phenotypic similarity, and variant quality. Starting from (1) the VCF file containing proband's variants, (2) the list of proband's phenotypes encoded in Human Phenotype Ontology terms, and optionally (3) the information about family members (if available), the "Suggested Diagnosis" ranks all the variants according to their machine learning prediction. This method significantly reduces the number of variants that need to be evaluated by geneticists by pinpointing causative variants in the very first positions of the prioritized list. Most importantly, our approach proved to be among the top performers within the CAGI6 Rare Genome Project Challenge, where it was able to rank the true causative variant among the first positions and, uniquely among all the challenge participants, increased the diagnostic yield of 12.5% by solving 2 undiagnosed cases.

Abstract 2275: Reducing artifactual somatic variant calls from formalin-fixed paraffin-embedded specimens by using DEEPOMICS FFPE, a bioinformatic approach based on deep neural networks

Abstract Formalin-fixed, paraffin-embedded (FFPE) tissue specimens are routinely used in pathological diagnosis. However, formalin induced artificial mutations make it challenging to accurately analyze next-generation sequencing data. Existing FFPE filtering tools tend to categorize all low-frequency variants as artifacts, making it difficult to identify true variants with low allele frequencies, compromising analysis of tumors with low tumor cell content. To address this issue, we developed DEEPOMICS FFPE, an AI model designed to distinguish true variants from artifacts. We gathered paired exome genome sequencing data from both fresh frozen and FFPE samples of 24 tumors from public sources. These data were used to train and validate the model, with a split of 70% for training and 30% for validation. The deep neural network model, comprising three hidden layers, was trained using input features derived from the outputs of the MuTect2 caller. We employed the SHapley Additive exPlanations algorithm to identify relevant features, and then fine-tuned them based on the training outcomes. We assessed the performance of the final DEEPOMICS FFPE model in comparison to existing models (MuTect filter, FFPolish, and SOBDetector) using well-defined test datasets. Our findings revealed 43 distinguishing characteristics for FFPE artifacts. Refining the quantification of these characteristics resulted in improved model performance. DEEPOMICS FFPE successfully filtered out 99.6% of artifacts while retaining 85.2% of true variants, achieving an F1-score of 87.8 in the validation dataset, surpassing the performance of existing tools. Furthermore, the model demonstrated consistent performance even for variants with low allele frequencies, exhibiting a specificity of 0.998, suggesting its capability to identify subclonal variants. Future enhancement of the model will include the ability to use other variant callers and analysis of whole genome sequencing data. This newly developed tool has potential applications in designing capture panels for personalized circulating tumor DNA assays and identifying candidate neoepitopes for personalized vaccine development. DEEPOMICS FFPE is freely accessible for research purposes on the web at http://deepomics.co.kr/ffpe. Citation Format: Dong-hyuk Heo, Inyoung Kim, Heejae Seo, Seong-Gwang Kim, Minji Kim, Jiin Park, Hongsil Park, Seungmo Kang, Juhee Kim, Soonmyung Paik, Seong-Eui Hong. Reducing artifactual somatic variant calls from formalin-fixed paraffin-embedded specimens by using DEEPOMICS FFPE, a bioinformatic approach based on deep neural networks [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2275.

Abstract 4926: Advancements in somatic variant calling from UG100 whole genome and whole exome sequencing data

Abstract Somatic variant calling involves the identification of genomic alterations that occur in somatic cells, requiring deep coverage to enable high sensitivity for low-frequency variants. Characterizing somatic variants across the entire genome therefore benefits from novel cost-efficient sequencing platforms, such as UG100. Here, we present optimization of variant calling tools for short and structural variants on WGS and WES data from UG100. For calling short variants, we optimized DeepVariant (DV) for somatic calling using data from matched tumor-normal sample pairs, improving both variant calling accuracy and pipeline running time (up to 10-fold). We defined the task of somatic variant calling as deciding if the pileup image containing reads from the tumor and normal samples represents a true somatic variant (vs a germline variant or artifact). The challenge of robust variant calling using deep learning models is exacerbated in somatic calling, where sequencing depth and coverage variability are typically high. Our optimized DV overcomes these challenges by several data sampling strategies. First, allele-frequency preserving down-sampling reduces randomness of read sub-sampling in high coverage regions. Second, alternative allele prioritization samples alt-allele supporting reads first allowing to call variants at very high coverage loci without sacrificing sensitivity and computational efficiency. Finally, a Panel-of-Normals based on targeted WES data provides an additional improvement of precision for this assay type. We used these strategies to train two models, one for tumor characterization using WGS (T/N coverage: 40x-150x/40x-100x), and one for deep WES (T/N coverage: &amp;gt;500x/&amp;gt;120x). We called variants on simulated tumors using the WGS model. For VAF&amp;gt;10% the model showed SNV recall &amp;gt;98% and indel recall &amp;gt;95% with false-positive rate of 0.2/Mb. For VAF range of 5-10%, indel recall was 67% and SNV recall was 86%. To demonstrate the utility of our somatic variant calling, we applied the models to call somatic variants from well characterized cancer cell lines: COLO829, HCC1395 and HCC1143. Results showed F1&amp;gt;90% for variants with VAF&amp;gt;10%. The WES model was used to reliably call variants at VAF&amp;gt;5% on simulated tumors with average SNV recall of 99% with precision &amp;gt;99% and indel recall &amp;gt;86% with precision &amp;gt;94%. To analyze structural and copy-number variations, we optimized the assembly engine of GRIDSS to enable fast calling of structural variations and demonstrate that Control-FREEC can be used to call copy number variants. SV calling on COLO829/COLO829BL achieved sensitivity &amp;gt;95%. In conclusion, our research highlights the utility of UG100 within the field of oncology, demonstrating its capacity for comprehensive and precise somatic variant detection, both on WGS and WES data. Citation Format: Doron Shem-Tov, Maya Levy, Gil Hornung, Ilya Soifer, Hila Benjamin, Ariel Jaimovich, Adam Blattler, William Brandler, Robert Sugar, Isaac Kinde, Omer Barad, Doron Lipson. Advancements in somatic variant calling from UG100 whole genome and whole exome sequencing data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4926.

Improving somatic exome sequencing performance by biological replicates

BackgroundNext-generation sequencing (NGS) technologies offer fast and inexpensive identification of DNA sequences. Somatic sequencing is among the primary applications of NGS, where acquired (non-inherited) variants are based on comparing diseased and healthy tissues from the same individual. Somatic mutations in genetic diseases such as cancer are tightly associated with genomic instability. Genomic instability increases heterogenity, complicating sequencing efforts further, a task already challenged by the presence of short reads and repetitions in human DNA. This leads to low concordance among studies and limits reproducibility. This limitation is a significant problem since identified mutations in somatic sequencing are major biomarkers for diagnosis and the primary input of targeted therapies. Benchmarking studies were conducted to assess the error rates and increase reproducibility. Unfortunately, the number of somatic benchmarking sets is very limited due to difficulties in validating true somatic variants. Moreover, most NGS benchmarking studies are based on relatively simpler germline (inherited) sequencing. Recently, a comprehensive somatic sequencing benchmarking set was published by Sequencing Quality Control Phase 2 (SEQC2). We chose this dataset for our experiments because it is a well-validated, cancer-focused dataset that includes many tumor/normal biological replicates. Our study has two primary goals. First goal is to determine how replicate-based consensus approaches can improve the accuracy of somatic variant detection systems. Second goal is to develop highly predictive machine learning (ML) models by employing replicate-based consensus variants as labels during the training phase.ResultsEnsemble approaches that combine alternative algorithms are relatively common; here, as an alternative, we study the performance enhancement potential of biological replicates. We first developed replicate-based consensus approaches that utilize the biological replicates available in this study to improve variant calling performance. Subsequently, we trained ML models using these biological replicates and achieved performance comparable to optimal ML models, those trained using high-confidence variants identified in advance.ConclusionsOur replicate-based consensus approach can be used to improve variant calling performance and develop efficient ML models. Given the relative ease of obtaining biological replicates, this strategy allows for the development of efficient ML models tailored to specific datasets or scenarios.

Abstract 909: Enhancing genomic analysis in cancer diagnostics: A machine learning approach for removing artifacts in FFPE specimens

Abstract Formalin-Fixed Paraffin-Embedded (FFPE) specimens, widely utilized in clinical cancer diagnostics, present significant challenges by introducing artifacts into genomic data. This study aimed to profile these FFPE-induced genomic alterations, with a particular focus on single-nucleotide variants (SNVs), small insertions and deletions (indels), and copy-number variations (CNVs), and to develop computational methods for filtering out such artifacts. Our primary focus was twofold: first, to comprehensively characterize the unique error profile of FFPE specimens observed through whole-genome sequencing (WGS), and second, to construct artifact classifiers and noise filters for SNV/indels and CNVs. We utilized machine learning (ML) and signal-processing techniques on a dataset of FFPE and matched fresh-frozen (FF) samples. The dataset of 52 FFPE-FF pairs were obtained from four different medical institutes and from various cancers including liver, breast, colon, stomach, and lung cancer, with varied FFPE archiving times. We also analyzed additional FFPE-only samples to refine our methods. Our methodology incorporated advanced computational approaches, including stacking ensemble, transfer learning, and wavelet transform, to enhance robustness and accuracy. The method's design was centered around the notion of not only achieving high performance in distinguishing true signals from FFPE-induced artifacts but also addressing the real-world challenges posed by the varying quality and conditions of clinical FFPE samples. In the analysis, we found peculiar patterns of FFPE-specific error profile, including well-known cytosine deamination and novel mutational signatures. The classifier, building on our findings, effectively differentiated true genomic variants from FFPE-induced artifacts for both SNVs and indels, demonstrating a sensitivity of 0.97, specificity of 0.87, and an F1 score of 0.94 for SNVs. For indels, it achieved a sensitivity of 0.91, specificity of 0.91, and an F1 score of 0.91. The CNV filter notably enhanced the signal-to-noise ratio (SNR) of CNV depth profiles, increasing it from 13dB to 17.5dB on average. Furthermore, we conducted evaluations on two critical measures in cancer and clinical genomics: homologous recombination deficiency (HRD) and tumor mutational burden (TMB), achieving post-filtering concordance rates of 0.99 for HRD—correctly identifying all 8 HRD-positive patients in our dataset—and 0.96 for SNV-based TMB and 0.87 for indel-based TMB. Additionally, a post hoc procedure for sensitive detection of cancer driver mutations resulted in concordance rates of 0.94 for SNVs, 0.91 for indels, and 0.95 for oncogene amplification. Taken together, our study advances FFPE WGS analysis in cancer diagnostics by effectively filtering artifacts and addressing challenges with older, degraded samples, enhancing clinical applicability. Citation Format: Joonoh Lim, Seongyeol Park, Won-Chul Lee, Ryul Kim, Sangmoon Lee, Jeong Seok Lee, Brian Baek-Lok Oh, Young Seok Ju. Enhancing genomic analysis in cancer diagnostics: A machine learning approach for removing artifacts in FFPE specimens [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 909.

Abstract 6252: Driver indel discovery and allelic imbalance in &amp;gt;9,000 tumor RNA-Seq samples from the Cancer Genome Atlas (TCGA)

Abstract High-throughput DNA sequencing technologies have enabled unbiased screening of genomic alterations such as single nucleotide variants (SNVs) and small insertions/deletions (indels). However, variant analysis using transcriptomic sequencing (RNA-seq) data has not become standard due to challenges in distinguishing true variants, in particular indels, from artifacts that can arise from RNA-seq mapping and library preparation. We have previously developed a tool, RNAIndel, which classifies indels as somatic, germline, or artifacts using a machine learning-based model. Here, we present enhanced somatic indel discovery by incorporating tumor RNA-Seq data with conventional paired tumor-normal DNA-Seq by re-analyzing the Cancer Genome Atlas (TCGA) data using RNAIndel. Our analytic process involves running RNAIndel on 9,101 TCGA tumor RNA-Seq samples across 33 cancer types hosted on Cancer Genomics Cloud by Seven Bridges Genomics (https://www.cancergenomicscloud.org) and comparing the result with the reference somatic indel set generated from paired tumor/normal whole-exome sequencing (WES) data using an ensemble caller developed by NCI Genomic Data Commons (GDC). The comparison validates the RNAIndel variants by matching to the corresponding WES dataset. We also use the GDC indel set to test their expression in RNA-Seq. Indel alignments across sequencing platforms were compared by indelPost, an algorithm that we developed, to perform indel realignment to overcome the differences in mapping algorithms (BWA vs. STAR) and read lengths (100-bp vs. 50-bp) in WES versus RNA-Seq data.The joint DNA/RNA indel analysis provides important insights into the expression profile of indels in known cancer driver genes versus non-drivers. First, indels in driver genes are more likely expressed than those in non-drivers (79% vs. 50%, p &amp;lt; 2.2 × 10-16). Second, mutant allele expression is frequently upregulated for truncating indels in tumor-suppressor genes (TSG) as allelic imbalance was detected in 6% of such variants in TSGs vs. 2% non-drivers. This may indicate a potential second hit leading to loss of the wild-type allele of in the TSGs. Third, allelic imbalance for in-frame indels was also frequent for oncogenes (20%) compared with the non-drivers (7%). The high expression likelihood of driver genes and the enriched indel allele expression for TSGs and oncogenes both support the use of RNA-Seq to enhance driver indel discovery. Indeed, our analysis reveals ~10% additional driver indels by RNA-Seq analysis which are absent from the GDC reference indel set. These indels affect prominent driver genes, most frequently CDKN2A, TP53 and ARID1A. Our experience argues for incorporating de novo indel analysis in RNA-Seq as a standard approach for future cancer genomic analysis. Citation Format: Kohei Hagiwara, Andrew Thrasher, Jinghui Zhang. Driver indel discovery and allelic imbalance in &amp;gt;9,000 tumor RNA-Seq samples from the Cancer Genome Atlas (TCGA) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6252.

Abstract 3677: Hybrid capture based sequencing from urinary cell free DNA from colorectal cancer patients

Abstract Background: The study of body fluids other than blood is gaining increasing attention in the field of liquid biopsy. Since urine offers a truly non-invasive sampling method, it is a particularly interesting specimen. Evidence suggests that urine cell-free DNA (ucfDNA) harbors information about renal and bladder cancer. However, not much is known about ucfDNA in colorectal cancer (CRC) patients. Therefore, we aimed to test the feasibility of a hybrid capture based NGS approach with and without preanalytical stabilization of urine samples. Methods: Urine samples were collected from 20 patients with metastatic CRC (mCRC), aliquoted and stored with or without a stabilizing agents (PAXgene Urine Liquid Biopsy Set and Streck urine Preserve) at room temperature. Urine cfDNA was isolated at the day of donation and after 3 days. Matched blood samples were additionally collected in PAXgene Blood ccfDNA Tubes (PreAnalytiX). cfDNA from blood and urine was isolated using the QIAsymphony platform (QIAGEN). cfDNA samples were analyzed using the AVENIO ctDNA Analysis Kit (Roche), a hybrid-capture based approach enriching for 17 clinically relevant genes. Results: While all stabilized ucfDNA yielded high quality libraries, library preparation failed in 66.7% of unstabilized urine samples, demonstrating that without stabilization ucfDNA rapidly degrades after urine donation. In native samples, in which sequencing data could be obtained, sequencing depth was significantly decreased compared to stabilized samples. For stabilized samples, sequence analysis revealed full concordance between urine and plasma for putative germline variants. However, when using less than 50ng of input material of ucfDNA, a low signal-to-noise ratio with a high number of false positive low level variants was observed. We therefore adjusted the limit of detection for ucfDNA to 0.5%, at which CRC-related mutations were detected in ucfDNA in one patient. Although these variants, were not observed in plasma cfDNA, they could be detected in all stabilized urine samples, which hints to true variants. Conclusion: Our data demonstrate that a hybrid-capture based analysis approach is feasible for ucfDNA form CRC patients and that urine may provide complementary information about a patient's tumor that may be missed in plasma. Yet, due to degradation and lower concentrations, immediate stabilization of urine after donation is required. Moreover, the limit of detection needs to be adjusted if low amounts of input DNA is available only. Citation Format: Anna Eberhard, Tina Moser, Leandra Ziegler, Georgios Vlachos, Isaac Lazzeri, Martina Loibner, Armin Gerger, Ellen Heitzer. Hybrid capture based sequencing from urinary cell free DNA from colorectal cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3677.

Improved detection of low-frequency within-host variants from deep sequencing: A case study with human papillomavirus.

High-coverage sequencing allows the study of variants occurring at low frequencies within samples, but is susceptible to false-positives caused by sequencing error. Ion Torrent has a very low single nucleotide variant (SNV) error rate and has been employed for the majority of human papillomavirus (HPV) whole genome sequences. However, benchmarking of intrahost SNVs (iSNVs) has been challenging, partly due to limitations imposed by the HPV life cycle. We address this problem by deep sequencing three replicates for each of 31 samples of HPV type 18 (HPV18). Errors, defined as iSNVs observed in only one of three replicates, are dominated by C→T (G→A) changes, independently of trinucleotide context. True iSNVs, defined as those observed in all three replicates, instead show a more diverse SNV type distribution, with particularly elevated C→T rates in CCG context (CCG→CTG; CGG→CAG) and C→A rates in ACG context (ACG→AAG; CGT→CTT). Characterization of true iSNVs allowed us to develop two methods for detecting true variants: (1) VCFgenie, a dynamic binomial filtering tool which uses each variant's allele count and coverage instead of fixed frequency cut-offs; and (2) a machine learning binary classifier which trains eXtreme Gradient Boosting models on variant features such as quality and trinucleotide context. Each approach outperforms fixed-cut-off filtering of iSNVs, and performance is enhanced when both are used together. Our results provide improved methods for identifying true iSNVs in within-host applications across sequencing platforms, specifically using HPV18 as a case study.

DEEPOMICS FFPE, a deep neural network model, identifies DNA sequencing artifacts from formalin fixed paraffin embedded tissue with high accuracy

Formalin-fixed, paraffin-embedded (FFPE) tissue specimens are routinely used in pathological diagnosis, but their large number of artifactual mutations complicate the evaluation of companion diagnostics and analysis of next-generation sequencing data. Identification of variants with low allele frequencies is challenging because existing FFPE filtering tools label all low-frequency variants as artifacts. To address this problem, we aimed to develop DEEPOMICS FFPE, an AI model that can classify a true variant from an artifact. Paired whole exome sequencing data from fresh frozen and FFPE samples from 24 tumors were obtained from public sources and used as training and validation sets at a ratio of 7:3. A deep neural network model with three hidden layers was trained with input features using outputs of the MuTect2 caller. Contributing features were identified using the SHapley Additive exPlanations algorithm and optimized based on training results. The performance of the final model (DEEPOMICS FFPE) was compared with those of existing models (MuTect filter, FFPolish, and SOBDetector) by using well-defined test datasets. We found 41 discriminating properties for FFPE artifacts. Optimization of property quantification improved the model performance. DEEPOMICS FFPE removed 99.6% of artifacts while maintaining 87.1% of true variants, with an F1-score of 88.3 in the entire dataset not used for training, which is significantly higher than those of existing tools. Its performance was maintained even for low-allele-fraction variants with a specificity of 0.995, suggesting that it can be used to identify subclonal variants. Different from existing methods, DEEPOMICS FFPE identified most of the sequencing artifacts in the FFPE samples while retaining more of true variants, including those of low allele frequencies. The newly developed tool DEEPOMICS FFPE may be useful in designing capture panels for personalized circulating tumor DNA assay and identifying candidate neoepitopes for personalized vaccine design. DEEPOMICS FFPE is freely available on the web (http://deepomics.co.kr/ffpe) for research.

A narrative review on rare types of pancreatic cancer: should they be treated as pancreatic ductal adenocarcinomas?

Pancreatic cancer is one of the deadliest malignancies in humans and it is expected to play a bigger part in cancer burden in the years to come. Pancreatic ductal adenocarcinoma (PDAC) represents 85% of all primary pancreatic malignancies. Recently, much attention has been given to PDAC, with significant advances in the understanding of the mechanisms underpinning disease initiation and progression, along with noticeable improvements in overall survival in both localized and metastatic settings. However, given their rarity, rare histological subtypes of pancreatic cancer have been underappreciated and are frequently treated as PDAC, even though they might present non-overlapping molecular alterations and clinical behavior. While some of these rare histological subtypes are true variants of PDAC that should be treated likewise, others represent separate clinicopathological entities, warranting a different therapeutic approach. In this review, we highlight clinical, pathological, and molecular aspects of rare histological types of pancreatic cancer, along with the currently available data to guide treatment decisions.

The diagnostic yield of exome sequencing in liver diseases from a curated gene panel

Exome sequencing (ES) has been used in a variety of clinical settings but there are limited data on its utility for diagnosis and/or prediction of monogenic liver diseases. We developed a curated list of 502 genes for monogenic disorders associated with liver phenotypes and analyzed ES data for these genes in 758 patients with chronic liver diseases (CLD). For comparison, we examined ES data in 7856 self-declared healthy controls (HC), and 2187 patients with chronic kidney disease (CKD). Candidate pathogenic (P) or likely pathogenic (LP) variants were initially identified in 19.9% of participants, most of which were attributable to previously reported pathogenic variants with implausibly high allele frequencies. After variant annotation and filtering based on population minor allele frequency (MAF ≤ 10–4 for dominant disorders and MAF ≤ 10–3 for recessive disorders), we detected a significant enrichment of P/LP variants in the CLD cohort compared to the HC cohort (X2 test OR 5.00, 95% CI 3.06–8.18, p value = 4.5e−12). A second-level manual annotation was necessary to capture true pathogenic variants that were removed by stringent allele frequency and quality filters. After these sequential steps, the diagnostic rate of monogenic disorders was 5.7% in the CLD cohort, attributable to P/LP variants in 25 genes. We also identified concordant liver disease phenotypes for 15/22 kidney disease patients with P/LP variants in liver genes, mostly associated with cystic liver disease phenotypes. Sequencing results had many implications for clinical management, including familial testing for early diagnosis and management, preventative screening for associated comorbidities, and in some cases for therapy. Exome sequencing provided a 5.7% diagnostic rate in CLD patients and required multiple rounds of review to reduce both false positive and false negative findings. The identification of concordant phenotypes in many patients with P/LP variants and no known liver disease also indicates a potential for predictive testing for selected monogenic liver disorders.

Improving fine-mapping by modeling infinitesimal effects.

Fine-mapping aims to identify causal genetic variants for phenotypes. Bayesian fine-mapping algorithms (for example, SuSiE, FINEMAP, ABF and COJO-ABF) are widely used, but assessing posterior probability calibration remains challenging in real data, where model misspecification probably exists, and true causal variants are unknown. We introduce replication failure rate (RFR), a metric to assess fine-mapping consistency by downsampling. SuSiE, FINEMAP and COJO-ABF show high RFR, indicating potential overconfidence in their output. Simulations reveal that nonsparse genetic architecture can lead to miscalibration, while imputation noise, nonuniform distribution of causal variants and quality control filters have minimal impact. Here we present SuSiE-inf and FINEMAP-inf, fine-mapping methods modeling infinitesimal effects alongside fewer larger causal effects. Our methods show improved calibration, RFR and functional enrichment, competitive recall and computational efficiency. Notably, using our methods' posterior effect sizes substantially increases polygenic risk score accuracy over SuSiE and FINEMAP. Our work improves causal variant identification for complex traits, a fundamental goal of human genetics.