A machine learning approach for somatic mutation discovery

Advancements and hurdles in the development of a vaccine for triple-negative breast cancer: A comprehensive review of multi-omics and immunomics strategies

Triple-Negative Breast Cancer (TNBC) presents a significant challenge in oncology due to its aggressive behavior and limited therapeutic options. This review explores the potential of immunotherapy, particularly vaccine-based approaches, in addressing TNBC. It delves into the role of immunoinformatics in creating effective vaccines against TNBC. The review first underscores the distinct attributes of TNBC and the importance of tumor antigens in vaccine development. It then elaborates on antigen detection techniques such as exome sequencing, HLA typing, and RNA sequencing, which are instrumental in identifying TNBC-specific antigens and selecting vaccine candidates. The discussion then shifts to the in-silico vaccine development process, encompassing antigen selection, epitope prediction, and rational vaccine design. This process merges computational simulations with immunological insights. The role of Artificial Intelligence (AI) in expediting the prediction of antigens and epitopes is also emphasized. The review concludes by encapsulating how Immunoinformatics can augment the design of TNBC vaccines, integrating tumor antigens, advanced detection methods, in-silico strategies, and AI-driven insights to advance TNBC immunotherapy. This could potentially pave the way for more targeted and efficacious treatments.

The mutational pattern of homologous recombination repair genes in urothelial carcinoma and its correlation with immunotherapeutic response

BACKGROUND

The mutational pattern of homologous recombination repair (HRR)-associated gene alterations in Chinese urothelial carcinoma (UC) necessitates comprehensive sequencing efforts, and the clinical implications of HRR gene mutations in UC remain to be elucidated.

MATERIALS AND METHODS

We delineated the mutational landscape of 343 Chinese UC patients from West China Hospital and 822 patients from The Cancer Genome Atlas (TCGA) using next-generation sequencing (NGS). Data from 182 metastatic UC patients from MSK-IMPACT cohort were used to assess the association between HRR mutations and immunotherapy efficacy. Comprehensive transcriptomic analysis was performed to explore the impact of HRR mutations on tumor immune microenvironment.

RESULTS

Among Chinese UC patients, 34% harbored HRR gene mutations, with BRCA2, ATM, BRCA1, CDK12, and RAD51C being the most prevalently mutated genes. Mutational signatures contributing to UC differed between patients with and without HRR mutations. Signature 22 for exposure to aristolochic acid was only observed in Chinese UC patients. The presence of HRR mutations was correlated with higher tumor mutational burden, neoantigen burden, and PD-L1 expression. Importantly, patients with HRR mutations exhibited significantly improved prognosis following immunotherapy compared to those without HRR mutations.

CONCLUSIONS

Our findings provide valuable insights into the genomic landscape of Chinese UC patients and underscore the molecular rationale for utilizing immunotherapy in UC patients with HRR mutations.

Validation of a Molecular Diagnostic Test for Circulating Tumor DNA by Next-Gen Sequencing

A modified version of the PGDx elioTM Plasma Resolve assay was validated as a laboratory-developed test (LDT) for clinical use in the Molecular Diagnostics Laboratory at Fox Chase Cancer Center. The test detects single nucleotide variants (SNVs) and small insertions and deletions (indels) in 33 target genes using fragmented genomic DNA extracted from plasma. The analytical performance of this assay was assessed with reference standard DNA and 29 samples from cancer patients and detected 66 SNVs and 23 indels. Using 50 ng of input DNA, the sensitivity was 95.5% to detect SNVs at 0.5% allele frequency, and the specificity was 92.3%. The sensitivity to detect indels at 1% allele frequency was 70.4%. A cutoff of 0.25% variant allele frequency (VAF) was set up for diagnostic reporting. An inter-laboratory study of concordance with an orthologous test resulted in a positive percent agreement (PPA) of 91.7%.

Comprehensive NGS profiling to enable detection of ALK gene rearrangements and MET amplifications in non-small cell lung cancer

Introduction

Next-generation sequencing (NGS) is currently widely used for biomarker studies and molecular profiling to identify concurrent alterations that can lead to the better characterization of a tumor's molecular landscape. However, further evaluation of technical aspects related to the detection of gene rearrangements and copy number alterations is warranted.

Methods

There were 12 ALK rearrangement-positive tumor specimens from patients with non-small cell lung cancer (NSCLC) previously detected via fluorescence in situ hybridization (FISH), immunohistochemistry (IHC), and an RNA-based NGS assay, and 26 MET high gene copy number (GCN) cases detected by FISH, selected for this retrospective study. All 38 pre-characterized cases were reassessed utilizing the PGDx™ elio™ tissue complete assay, a 505 gene targeted NGS panel, to evaluate concordance with these conventional diagnostic techniques.

Results

The detection of ALK rearrangements using the DNA-based NGS assay demonstrated excellent sensitivity with the added benefit of characterizing gene fusion partners and genomic breakpoints. MET copy number alterations were also detected; however, some discordances were observed likely attributed to differences in algorithm, reporting thresholds and gene copy number state. TMB was also assessed by the assay and correlated to the presence of NSCLC driver alterations and was found to be significantly lower in cases with NGS-confirmed canonical driver mutations compared with those without (p=0.0019).

Discussion

Overall, this study validates NGS as an accurate approach for detecting structural variants while also highlighting the need for further optimization to enable harmonization across methodologies for amplifications.

ctDNA response after pembrolizumab in non-small cell lung cancer: phase 2 adaptive trial results

Circulating tumor DNA (ctDNA) has shown promise in capturing primary resistance to immunotherapy. BR.36 is a multi-center, randomized, ctDNA-directed, phase 2 trial of molecular response-adaptive immuno-chemotherapy for patients with lung cancer. In the first of two independent stages, 50 patients with advanced non-small cell lung cancer received pembrolizumab as standard of care. The primary objectives of stage 1 were to ascertain ctDNA response and determine optimal timing and concordance with radiologic Response Evaluation Criteria in Solid Tumors (RECIST) response. Secondary endpoints included the evaluation of time to ctDNA response and correlation with progression-free and overall survival. Maximal mutant allele fraction clearance at the third cycle of pembrolizumab signified molecular response (mR). The trial met its primary endpoint, with a sensitivity of ctDNA response for RECIST response of 82% (90% confidence interval (CI): 52-97%) and a specificity of 75% (90% CI: 56.5-88.5%). Median time to ctDNA response was 2.1 months (90% CI: 1.5-2.6), and patients with mR attained longer progression-free survival (5.03 months versus 2.6 months) and overall survival (not reached versus 7.23 months). These findings are incorporated into the ctDNA-driven interventional molecular response-adaptive second stage of the BR.36 trial in which patients at risk of progression are randomized to treatment intensification or continuation of therapy. ClinicalTrials.gov ID: NCT04093167 .

Validation of the Labcorp Plasma Focus Test to Facilitate Precision Oncology Through Cell-Free DNA Genomic Profiling of Solid Tumors

Genomic profiling is critical for precision oncology to guide treatment decisions. Liquid biopsy testing is a complementary approach to tissue testing, particularly when tissue is not readily available. The Labcorp Plasma Focus test is a circulating cell-free DNA genomic profiling test that identifies actionable variants in solid cancers, including non-small-cell lung, colorectal, melanoma, breast, esophageal, gastroesophageal junction, and gastric cancers. This study highlights the analytical validation of the test, including accuracy compared with orthogonal methods, as well as sensitivity, specificity, precision, reproducibility, and repeatability. Concordance with orthogonal methods showed percent positive agreement of 98.7%, 89.3%, and 96.2% for single nucleotide variants (SNVs), insertion/deletions (indels), and copy number amplifications (CNAs), respectively, and 100.0% for translocations and microsatellite instability (MSI). Analytical sensitivity revealed a median limit of detection of 0.7% and 0.6% for SNVs and indels, 1.4-fold for CNAs, 0.5% variant allele frequency for translocations, and 0.6% for MSI. Specificity was >99% for SNVs/indels and 100% for CNAs, translocations, and MSI. Average positive agreement from precision, reproducibility, and repeatability experiments was 97.5% and 88.9% for SNVs/indels and CNAs, and 100% for translocations and MSI. Taken together, these data show that the Labcorp Plasma Focus test is a highly accurate, sensitive, and specific approach for cell-free DNA genomic profiling to supplement tissue testing and inform treatment decisions.

A systematic review on machine learning approaches in the diagnosis and prognosis of rare genetic diseases

BACKGROUND

The diagnosis of rare genetic diseases is often challenging due to the complexity of the genetic underpinnings of these conditions and the limited availability of diagnostic tools. Machine learning (ML) algorithms have the potential to improve the accuracy and speed of diagnosis by analyzing large amounts of genomic data and identifying complex multiallelic patterns that may be associated with specific diseases. In this systematic review, we aimed to identify the methodological trends and the ML application areas in rare genetic diseases.

METHODS

We performed a systematic review of the literature following the PRISMA guidelines to search studies that used ML approaches to enhance the diagnosis of rare genetic diseases. Studies that used DNA-based sequencing data and a variety of ML algorithms were included, summarized, and analyzed using bibliometric methods, visualization tools, and a feature co-occurrence analysis.

FINDINGS

Our search identified 22 studies that met the inclusion criteria. We found that exome sequencing was the most frequently used sequencing technology (59%), and rare neoplastic diseases were the most prevalent disease scenario (59%). In rare neoplasms, the most frequent applications of ML models were the differential diagnosis or stratification of patients (38.5%) and the identification of somatic mutations (30.8%). In other rare diseases, the most frequent goals were the prioritization of rare variants or genes (55.5%) and the identification of biallelic or digenic inheritance (33.3%). The most employed method was the random forest algorithm (54.5%). In addition, the features of the datasets needed for training these algorithms were distinctive depending on the goal pursued, including the mutational load in each gene for the differential diagnosis of patients, or the combination of genotype features and sequence-derived features (such as GC-content) for the identification of somatic mutations.

CONCLUSIONS

ML algorithms based on sequencing data are mainly used for the diagnosis of rare neoplastic diseases, with random forest being the most common approach. We identified key features in the datasets used for training these ML models according to the objective pursued. These features can support the development of future ML models in the diagnosis of rare genetic diseases.

Data mining on identifying diagnosis and prognosis biomarkers in head and neck squamous carcinoma

Head and neck squamous carcinoma (HNSC) induces high cancer-related death worldwide. The biomarker screening on diagnosis and prognosis is of great importance. This research is aimed to explore the specific diagnostic and prognostic biomarkers for HNSC through bioinformatics analysis. The mutation and dysregulation data were acquired from UCSC Xena and TCGA databases. The top ten genes with mutation frequency in HNSC were TP53 (66%), TTN (35%), FAT1 (21%), CDKN2A (20%), MUC16 (17%), CSMD3 (16%), PIK3CA (16%), NOTCH1 (16%), SYNE1 (15%), LRP1B (14%). A total of 1,060 DEGs were identified, with 396 up-regulated and 665 downregulated in HNSC patients. Patients with lower expression of ACTN2 (P = 0.039, HR = 1.3), MYH1 (P = 0.005, HR = 1.5), MYH2 (P = 0.035, HR = 1.3), MYH7 (P = 0.053, HR = 1.3), and NEB (P = 0.0043, HR = 1.5) exhibit longer overall survival time in HNSC patients. The main DEGs were further analyzed by pan-cancer expression and immune cell infiltration analyses. MYH1, MYH2, and MYH7 were dysregulated in the cancers. Compared with HNSC, their expression levels are lower in the other types of cancers. MYH1, MYH2, and MYH7 were expected to be the specific diagnostic and prognostic molecular biomarkers of HNSC. All five DEGs have a significant positive correlation with CD4+T cells and macrophages.

Universal cutoff for tumor mutational burden in predicting the efficacy of anti-PD-(L)1 therapy for advanced cancers

Background: The US Food and Drug Administration (FDA)'s tumor-agnostic approval of pembrolizumab in high tumor mutational burden (TMB-high, i.e., TMB≥10 mut/Mb) cases, based on the data from KEYNOTE-158, has raised considerable concerns among the immuno-oncology community. This study aims to statistically infer the optimal universal cutoff in defining TMB-high that is predictive of the efficacy of anti-PD-(L) 1 therapy in advanced solid tumors. Methods: We integrated MSK-IMPACT TMB data from a public cohort and the objective response rate (ORR) for anti-PD-(L) 1 monotherapy across diverse cancer types in published trials. The optimal TMB cutoff was determined by varying the universal cutoff to define TMB-high across cancer types and examining the cancer-level correlation between objective response rate and the proportion of TMB-high cases. The utility of this cutoff in predicting overall survival (OS) benefits from anti-PD-(L) 1 therapy was then evaluated in a validation cohort of advanced cancers with coupled MSK-IMPACT TMB and OS data. In silico analysis of whole-exome sequencing data from The Cancer Genome Atlas was further employed to assess the generalizability of the identified cutoff among panels comprising several hundred genes. Results: The cancer type-level analysis identified 10 mut/Mb as the optimal cutoff for MSK-IMPACT in defining TMB-high, with the corresponding TMB-high (TMB≥10 mut/Mb) percentage strongly correlated with ORR for PD-(L) 1 blockade across cancer types [correlation coefficient, 0.72 (95% CI, 0.45-0.88)]. This cutoff was also the optimum in defining TMB-high (via MSK-IMPACT) when predicting OS benefits from anti-PD-(L) 1 therapy in the validation cohort. In this cohort, TMB≥10 mut/Mb was associated with significantly improved OS (hazard ratio, 0.58 [95% CI, 0.48-0.71]; p < 0.001). Moreover, in silico analyses revealed excellent agreement of TMB≥10 mut/Mb cases between MSK-IMPACT and the FDA-approved panels and between MSK-IMPACT and various randomly sampled panels. Conclusion: Our study demonstrates that 10 mut/Mb is the optimal, universal cutoff for TMB-high that guides the clinical application of anti-PD-(L) 1 therapy for advanced solid tumors. It also provides rigorous evidence beyond KEYNOTE-158 for the utility of TMB≥10 mut/Mb in predicting the efficacy of PD-(L) 1 blockade in broader settings, which could help to mitigate the challenges in embracing the tumor-agnostic approval of pembrolizumab in TMB-high cases.

Metastatic Colorectal Cancer Treatment Response Evaluation by Ultra-Deep Sequencing of Cell-Free DNA and Matched White Blood Cells

PURPOSE

Circulating tumor DNA (ctDNA) has the potential to guide therapy selection and monitor treatment response in patients with metastatic cancer. However, germline and clonal hematopoiesis-associated alterations can confound identification of tumor-specific mutations in cell-free DNA (cfDNA), often requiring additional sequencing of tumor tissue. The current study assessed whether ctDNA-based treatment response monitoring could be performed in a tumor tissue-independent manner by combining ultra-deep targeted sequencing analyses of cfDNA with patient-matched white blood cell (WBC)-derived DNA.

EXPERIMENTAL DESIGN

In total, 183 cfDNA and 49 WBC samples, along with 28 tissue samples, from 52 patients with metastatic colorectal cancer participating in the prospective phase III CAIRO5 clinical trial were analyzed using an ultra-deep targeted sequencing liquid biopsy assay.

RESULTS

The combined cfDNA and WBC analysis prevented false-positives due to germline or hematopoietic variants in 40% of patients. Patient-matched tumor tissue sequencing did not provide additional information. Longitudinal analyses of ctDNA were more predictive of overall survival than standard-of-care radiological response evaluation. ctDNA mutations related to primary or acquired resistance to panitumumab were identified in 42% of patients.

CONCLUSIONS

Accurate calling of ctDNA mutations for treatment response monitoring is feasible in a tumor tissue-independent manner by combined cfDNA and patient-matched WBC genomic DNA analysis. This tissue biopsy-independent approach simplifies sample logistics and facilitates the application of liquid biopsy ctDNA testing for evaluation of emerging therapy resistance, opening new avenues for early adaptation of treatment regimens.

Review of diagnosis, differential diagnosis, and management of retroperitoneal lymphangioma

Lymphatic malformation (LM) is the currently preferred term for what was previously known as lymphangioma. Retroperitoneal LMs are extremely rare, benign, cystic masses that arise from lymphatic vessels. They can be challenging to diagnose because they resemble other retroperitoneal cystic tumors. The development of treatment strategies for rare diseases, including retroperitoneal LM, requires the acquisition of new knowledge to enhance our understanding of the disease progression. Therefore, we present an update regarding fundamental and advanced issues associated with retroperitoneal LM. This review describes the epidemiology, histopathology, biomedicine, clinical manifestations, radiological features, differential diagnosis, and management of this lesion.

DeepSom: a CNN-based approach to somatic variant calling in WGS samples without a matched normal

MOTIVATION

Somatic mutations are usually called by analyzing the DNA sequence of a tumor sample in conjunction with a matched normal. However, a matched normal is not always available, for instance, in retrospective analysis or diagnostic settings. For such cases, tumor-only somatic variant calling tools need to be designed. Previously proposed approaches demonstrate inferior performance on whole-genome sequencing (WGS) samples.

RESULTS

We present the convolutional neural network-based approach called DeepSom for detecting somatic single nucleotide polymorphism and short insertion and deletion variants in tumor WGS samples without a matched normal. We validate DeepSom by reporting its performance on five different cancer datasets. We also demonstrate that on WGS samples DeepSom outperforms previously proposed methods for tumor-only somatic variant calling.

AVAILABILITY AND IMPLEMENTATION

DeepSom is available as a GitHub repository at https://github.com/heiniglab/DeepSom.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Artificial intelligence applied in neoantigen identification facilitates personalized cancer immunotherapy

The field of cancer neoantigen investigation has developed swiftly in the past decade. Predicting novel and true neoantigens derived from large multi-omics data became difficult but critical challenges. The rise of Artificial Intelligence (AI) or Machine Learning (ML) in biomedicine application has brought benefits to strengthen the current computational pipeline for neoantigen prediction. ML algorithms offer powerful tools to recognize the multidimensional nature of the omics data and therefore extract the key neoantigen features enabling a successful discovery of new neoantigens. The present review aims to outline the significant technology progress of machine learning approaches, especially the newly deep learning tools and pipelines, that were recently applied in neoantigen prediction. In this review article, we summarize the current state-of-the-art tools developed to predict neoantigens. The standard workflow includes calling genetic variants in paired tumor and blood samples, and rating the binding affinity between mutated peptide, MHC (I and II) and T cell receptor (TCR), followed by characterizing the immunogenicity of tumor epitopes. More specifically, we highlight the outstanding feature extraction tools and multi-layer neural network architectures in typical ML models. It is noted that more integrated neoantigen-predicting pipelines are constructed with hybrid or combined ML algorithms instead of conventional machine learning models. In addition, the trends and challenges in further optimizing and integrating the existing pipelines are discussed.

Fast, accurate, and racially unbiased pan-cancer tumor-only variant calling with tabular machine learning

Accurately identifying somatic mutations is essential for precision oncology and crucial for calculating tumor-mutational burden (TMB), an important predictor of response to immunotherapy. For tumor-only variant calling (i.e., when the cancer biopsy but not the patient's normal tissue sample is sequenced), accurately distinguishing somatic mutations from germline variants is a challenging problem that, when unaddressed, results in unreliable, biased, and inflated TMB estimates. Here, we apply machine learning to the task of somatic vs germline classification in tumor-only solid tumor samples using TabNet, XGBoost, and LightGBM, three machine-learning models for tabular data. We constructed a training set for supervised classification using features derived exclusively from tumor-only variant calling and drawing somatic and germline truth labels from an independent pipeline using the patient-matched normal samples. All three trained models achieved state-of-the-art performance on two holdout test datasets: a TCGA dataset including sarcoma, breast adenocarcinoma, and endometrial carcinoma samples (AUC > 94%), and a metastatic melanoma dataset (AUC > 85%). Concordance between matched-normal and tumor-only TMB improves from R2 = 0.006 to 0.71-0.76 with the addition of a machine-learning classifier, with LightGBM performing best. Notably, these machine-learning models generalize across cancer subtypes and capture kits with a call rate of 100%. We reproduce the recent finding that tumor-only TMB estimates for Black patients are extremely inflated relative to that of white patients due to the racial biases of germline databases. We show that our approach with XGBoost and LightGBM eliminates this significant racial bias in tumor-only variant calling.

Artificial intelligence in cancer research and precision medicine: Applications, limitations and priorities to drive transformation in the delivery of equitable and unbiased care

Artificial intelligence (AI) has experienced explosive growth in oncology and related specialties in recent years. The improved expertise in data capture, the increased capacity for data aggregation and analytic power, along with decreasing costs of genome sequencing and related biologic "omics", set the foundation and need for novel tools that can meaningfully process these data from multiple sources and of varying types. These advances provide value across biomedical discovery, diagnosis, prognosis, treatment, and prevention, in a multimodal fashion. However, while big data and AI tools have already revolutionized many fields, medicine has partially lagged due to its complexity and multi-dimensionality, leading to technical challenges in developing and validating solutions that generalize to diverse populations. Indeed, inner biases and miseducation of algorithms, in view of their implementation in daily clinical practice, are increasingly relevant concerns; critically, it is possible for AI to mirror the unconscious biases of the humans who generated these algorithms. Therefore, to avoid worsening existing health disparities, it is critical to employ a thoughtful, transparent, and inclusive approach that involves addressing bias in algorithm design and implementation along the cancer care continuum. In this review, a broad landscape of major applications of AI in cancer care is provided, with a focus on cancer research and precision medicine. Major challenges posed by the implementation of AI in the clinical setting will be discussed. Potentially feasible solutions for mitigating bias are provided, in the light of promoting cancer health equity.

Integration of artificial intelligence and precision oncology in Latin America

Next-generation medicine encompasses different concepts related to healthcare models and technological developments. In Latin America and the Caribbean, healthcare systems are quite different between countries, and cancer control is known to be insufficient and inefficient considering socioeconomically discrepancies. Despite advancements in knowledge about the biology of different oncological diseases, the disease remains a challenge in terms of diagnosis, treatment, and prognosis for clinicians and researchers. With the development of molecular biology, better diagnosis methods, and therapeutic tools in the last years, artificial intelligence (AI) has become important, because it could improve different clinical scenarios: predicting clinically relevant parameters, cancer diagnosis, cancer research, and accelerating the growth of personalized medicine. The incorporation of AI represents an important challenge in terms of diagnosis, treatment, and prognosis for clinicians and researchers in cancer care. Therefore, some studies about AI in Latin America and the Caribbean are being conducted with the aim to improve the performance of AI in those countries. This review introduces AI in cancer care in Latin America and the Caribbean, and the advantages and promising results that it has shown in this socio-demographic context.

Sex-specific differences in immunogenomic features of response to immune checkpoint blockade

Introduction

The magnitude of response to immune checkpoint inhibitor (ICI) therapy may be sex-dependent, as females have lower response rates and decreased survival after ICI monotherapy. The mechanisms underlying this sex dimorphism in ICI response are unknown, and may be related to sex-driven differences in the immunogenomic landscape of tumors that shape anti-tumor immune responses in the context of therapy.

Methods

To investigate the association of immunogenic mutations with HLA haplotypes, we leveraged whole exome sequence data and HLA genotypes from 482 non-small cell lung cancer (NSCLC) tumors from The Cancer Genome Atlas (TCGA). To explore sex-specific genomic features linked with ICI response, we analyzed whole exome sequence data from patients with NSCLC treated with ICI. Tumor mutational burden (TMB), HLA class I and II restricted immunogenic missense mutation (IMM) load, and mutational smoking signature were defined for each tumor. IMM load was combined with HLA class I and II haplotypes and correlated with therapeutic response and survival following ICI treatment. We examined rates of durable clinical benefit (DCB) for at least six months from ICI treatment initiation. Findings were validated utilizing whole exome sequence data from an independent cohort of ICI treated NSCLC.

Results

Analysis of whole exome sequence data from NSCLC tumors of females and males revealed that germline HLA class II diversity (≥9 unique HLA alleles) was associated with higher tumor class II IMM load in females (p=0.01) and not in males (p=0.64). Similarly, in tumors of female patients, somatic HLA class II loss of heterozygosity was associated with increased IMM load (p=0.01) while this association was not observed in tumors in males (p=0.20). In females, TMB (p=0.005), class I IMM load (p=0.005), class II IMM load (p=0.004), and mutational smoking signature (p<0.001) were significantly higher in tumors responding to ICI as compared to non-responding tumors. In contrast, among males, there was no significant association between DCB and any of these features. When IMM was considered in the context of HLA zygosity, high MHC-II restricted IMM load and high HLA class II diversity was significantly associated with overall survival in males (p=0.017).

Conclusions

Inherent sex-driven differences in immune surveillance affect the immunogenomic determinants of response to ICI and likely mediate the dimorphic outcomes with ICI therapy. Deeper understanding of the selective pressures and mechanisms of immune escape in tumors in males and females can inform patient selection strategies and can be utilized to further hone immunotherapy approaches in cancer.

Machine learning approaches for biomolecular, biophysical, and biomaterials research

A fluent conversation with a virtual assistant, person-tailored news feeds, and deep-fake images created within seconds-all those things that have been unthinkable for a long time are now a part of our everyday lives. What these examples have in common is that they are realized by different means of machine learning (ML), a technology that has fundamentally changed many aspects of the modern world. The possibility to process enormous amount of data in multi-hierarchical, digital constructs has paved the way not only for creating intelligent systems but also for obtaining surprising new insight into many scientific problems. However, in the different areas of biosciences, which typically rely heavily on the collection of time-consuming experimental data, applying ML methods is a bit more challenging: Here, difficulties can arise from small datasets and the inherent, broad variability, and complexity associated with studying biological objects and phenomena. In this Review, we give an overview of commonly used ML algorithms (which are often referred to as "machines") and learning strategies as well as their applications in different bio-disciplines such as molecular biology, drug development, biophysics, and biomaterials science. We highlight how selected research questions from those fields were successfully translated into machine readable formats, discuss typical problems that can arise in this context, and provide an overview of how to resolve those encountered difficulties.

Automated next-generation profiling of genomic alterations in human cancers

The lack of validated, distributed comprehensive genomic profiling assays for patients with cancer inhibits access to precision oncology treatment. To address this, we describe elio tissue complete, which has been FDA-cleared for examination of 505 cancer-related genes. Independent analyses of clinically and biologically relevant sequence changes across 170 clinical tumor samples using MSK-IMPACT, FoundationOne, and PCR-based methods reveals a positive percent agreement of >97%. We observe high concordance with whole-exome sequencing for evaluation of tumor mutational burden for 307 solid tumors (Pearson r = 0.95) and comparison of the elio tissue complete microsatellite instability detection approach with an independent PCR assay for 223 samples displays a positive percent agreement of 99%. Finally, evaluation of amplifications and translocations against DNA- and RNA-based approaches exhibits >98% negative percent agreement and positive percent agreement of 86% and 82%, respectively. These methods provide an approach for pan-solid tumor comprehensive genomic profiling with high analytical performance.

Detection of oncogenic and clinically actionable mutations in cancer genomes critically depends on variant calling tools

Motivation

The analysis of cancer genomes provides fundamental information about its etiology, the processes driving cell transformation or potential treatments. While researchers and clinicians are often only interested in the identification of oncogenic mutations, actionable variants or mutational signatures, the first crucial step in the analysis of any tumor genome is the identification of somatic variants in cancer cells (i.e. those that have been acquired during their evolution). For that purpose, a wide range of computational tools have been developed in recent years to detect somatic mutations in sequencing data from tumor samples. While there have been some efforts to benchmark somatic variant calling tools and strategies, the extent to which variant calling decisions impact the results of downstream analyses of tumor genomes remains unknown.

Results

Here, we quantify the impact of variant calling decisions by comparing the results obtained in three important analyses of cancer genomics data (identification of cancer driver genes, quantification of mutational signatures and detection of clinically actionable variants) when changing the somatic variant caller (MuSE, MuTect2, SomaticSniper and VarScan2) or the strategy to combine them (Consensus of two, Consensus of three and Union) across all 33 cancer types from The Cancer Genome Atlas. Our results show that variant calling decisions have a significant impact on these analyses, creating important differences that could even impact treatment decisions for some patients. Moreover, the Consensus of three calling strategy to combine the output of multiple variant calling tools, a very widely used strategy by the research community, can lead to the loss of some cancer driver genes and actionable mutations. Overall, our results highlight the limitations of widespread practices within the cancer genomics community and point to important differences in critical analyses of tumor sequencing data depending on variant calling, affecting even the identification of clinically actionable variants.

Availability and implementation

Code is available at https://github.com/carlosgarciaprieto/VariantCallingClinicalBenchmark.

Supplementary information

Supplementary data are available at Bioinformatics online.

Computational analysis of cancer genome sequencing data

Distilling biologically meaningful information from cancer genome sequencing data requires comprehensive identification of somatic alterations using rigorous computational methods. As the amount and complexity of sequencing data have increased, so has the number of tools for analysing them. Here, we describe the main steps involved in the bioinformatic analysis of cancer genomes, review key algorithmic developments and highlight popular tools and emerging technologies. These tools include those that identify point mutations, copy number alterations, structural variations and mutational signatures in cancer genomes. We also discuss issues in experimental design, the strengths and limitations of sequencing modalities and methodological challenges for the future.

EPIMUTESTR: a nearest neighbor machine learning approach to predict cancer driver genes from the evolutionary action of coding variants

Discovering rare cancer driver genes is difficult because their mutational frequency is too low for statistical detection by computational methods. EPIMUTESTR is an integrative nearest-neighbor machine learning algorithm that identifies such marginal genes by modeling the fitness of their mutations with the phylogenetic Evolutionary Action (EA) score. Over cohorts of sequenced patients from The Cancer Genome Atlas representing 33 tumor types, EPIMUTESTR detected 214 previously inferred cancer driver genes and 137 new candidates never identified computationally before of which seven genes are supported in the COSMIC Cancer Gene Census. EPIMUTESTR achieved better robustness and specificity than existing methods in a number of benchmark methods and datasets.

Somatic Genomic Testing in Patients With Metastatic or Advanced Cancer: ASCO Provisional Clinical Opinion

PURPOSE

An ASCO provisional clinical opinion offers timely clinical direction to ASCO's membership following publication or presentation of potentially practice-changing data from major studies. This provisional clinical opinion addresses the appropriate use of tumor genomic testing in patients with metastatic or advanced solid tumors.

CLINICAL CONTEXT

An increasing number of therapies are approved to treat cancers harboring specific genomic biomarkers. However, there is a lack of clarity as to when tumor genomic sequencing should be ordered, what type of assays should be performed, and how to interpret the results for treatment selection.

PROVISIONAL CLINICAL OPINION

Patients with metastatic or advanced cancer should undergo genomic sequencing in a certified laboratory if the presence of one or more specific genomic alterations has regulatory approval as biomarkers to guide the use of or exclusion from certain treatments for their disease. Multigene panel-based assays should be used if more than one biomarker-linked therapy is approved for the patient's disease. Site-agnostic approvals for any cancer with a high tumor mutation burden, mismatch repair deficiency, or neurotrophic tyrosine receptor kinase (NTRK) fusions provide a rationale for genomic testing for all solid tumors. Multigene testing may also assist in treatment selection by identifying additional targets when there are few or no genotype-based therapy approvals for the patient's disease. For treatment planning, the clinician should consider the functional impact of the targeted alteration and expected efficacy of genomic biomarker-linked options relative to other approved or investigational treatments.Additional information is available at www.asco.org/assays-and-predictive-markers-guidelines.

mitoDataclean: A machine learning approach for the accurate identification of cross-contamination-derived tumor mitochondrial DNA mutations

Next-generation sequencing (NGS) of mitochondrial DNA (mtDNA) has widespread applications in aging and cancer studies. However, cross-contamination of mtDNA constitutes a major concern. Previous methods for the detection of mtDNA contamination mainly focus on haplogroup-level phylogeny, but neglect haplotype-level differences, leading to limited sensitivity and accuracy. In this study, we present mitoDataclean, a random-forest-based machine learning package for accurate identification of cross-contamination, evaluation of contamination levels and detection of contamination-derived variants in mtDNA NGS data. Comprehensive optimization of mitoDataclean revealed that training simulation with mixtures of small haplogroup distance and low polymorphic difference was critical for optimal modeling. Compared with existing methods, mitoDataclean exhibited significantly improved sensitivity and accuracy for the detection of sample contamination in simulated data. In addition, mitoDataclean achieved area under the curve values of 0.91 and 0.97 for discerning genuine and contamination-derived mtDNA variants in a simulated Western dataset and private sequencing contamination data, respectively, suggesting that this tool may be applicable for different populations and samples with different sources of contamination. Finally, mitoDataclean was further evaluated in several private and public datasets and showed a robust ability for contamination detection. Altogether, our study demonstrates that mitoDataclean may be used for accurate detection of contaminated samples and contamination-derived variants in mtDNA NGS data. This article is protected by copyright. All rights reserved.

Aligning tumor mutational burden (TMB) quantification across diagnostic platforms: phase II of the Friends of Cancer Research TMB Harmonization Project

BACKGROUND

Tumor mutational burden (TMB) measurements aid in identifying patients who are likely to benefit from immunotherapy; however, there is empirical variability across panel assays and factors contributing to this variability have not been comprehensively investigated. Identifying sources of variability can help facilitate comparability across different panel assays, which may aid in broader adoption of panel assays and development of clinical applications.

MATERIALS AND METHODS

Twenty-nine tumor samples and 10 human-derived cell lines were processed and distributed to 16 laboratories; each used their own bioinformatics pipelines to calculate TMB and compare to whole exome results. Additionally, theoretical positive percent agreement (PPA) and negative percent agreement (NPA) of TMB were estimated. The impact of filtering pathogenic and germline variants on TMB estimates was assessed. Calibration curves specific to each panel assay were developed to facilitate translation of panel TMB values to whole exome sequencing (WES) TMB values.

RESULTS

Panel sizes >667 Kb are necessary to maintain adequate PPA and NPA for calling TMB high versus TMB low across the range of cut-offs used in practice. Failure to filter out pathogenic variants when estimating panel TMB resulted in overestimating TMB relative to WES for all assays. Filtering out potential germline variants at >0% population minor allele frequency resulted in the strongest correlation to WES TMB. Application of a calibration approach derived from The Cancer Genome Atlas data, tailored to each panel assay, reduced the spread of panel TMB values around the WES TMB as reflected in lower root mean squared error (RMSE) for 26/29 (90%) of the clinical samples.

CONCLUSIONS

Estimation of TMB varies across different panels, with panel size, gene content, and bioinformatics pipelines contributing to empirical variability. Statistical calibration can achieve more consistent results across panels and allows for comparison of TMB values across various panel assays. To promote reproducibility and comparability across assays, a software tool was developed and made publicly available.

Machine learning random forest for predicting oncosomatic variant NGS analysis

Since 2017, we have used IonTorrent NGS platform in our hospital to diagnose and treat cancer. Analyzing variants at each run requires considerable time, and we are still struggling with some variants that appear correct on the metrics at first, but are found to be negative upon further investigation. Can any machine learning algorithm (ML) help us classify NGS variants? This has led us to investigate which ML can fit our NGS data and to develop a tool that can be routinely implemented to help biologists. Currently, one of the greatest challenges in medicine is processing a significant quantity of data. This is particularly true in molecular biology with the advantage of next-generation sequencing (NGS) for profiling and identifying molecular tumors and their treatment. In addition to bioinformatics pipelines, artificial intelligence (AI) can be valuable in helping to analyze mutation variants. Generating sequencing data from patient DNA samples has become easy to perform in clinical trials. However, analyzing the massive quantities of genomic or transcriptomic data and extracting the key biomarkers associated with a clinical response to a specific therapy requires a formidable combination of scientific expertise, biomolecular skills and a panel of bioinformatic and biostatistic tools, in which artificial intelligence is now successful in developing future routine diagnostics. However, cancer genome complexity and technical artifacts make identifying real variants challenging. We present a machine learning method for classifying pathogenic single nucleotide variants (SNVs), single nucleotide polymorphisms (SNPs), multiple nucleotide variants (MNVs), insertions, and deletions detected by NGS from different types of tumor specimens, such as: colorectal, melanoma, lung and glioma cancer. We compared our NGS data to different machine learning algorithms using the k-fold cross-validation method and to neural networks (deep learning) to measure the performance of the different ML algorithms and determine which one is a valid model for confirming NGS variant calls in cancer diagnosis. We trained our machine learning with 70% of our data samples, extracted from our local database (our data structure had 7 parameters: chromosome, position, exon, variant allele frequency, minor allele frequency, coverage and protein description) and validated it with the 30% remaining data. The model offering the best accuracy was chosen and implemented in the NGS analysis routine. Artificial intelligence was developed with the R script language version 3.6.0. We trained our model on 70% of 102,011 variants. Our best error rate (0.22%) was found with random forest machine learning (ntree = 500 and mtry = 4), with an AUC of 0.99. Neural networks achieved some good scores. The final trained model with the neural network achieved an accuracy of 98% and an ROC-AUC of 0.99 with validation data. We tested our RF model to interpret more than 2000 variants from our NGS database: 20 variants were misclassified (error rate < 1%). The errors were nomenclature problems and false positives. After adding false positives to our training database and implementing our RF model routinely, our error rate was always < 0.5%. The RF model shows excellent results for oncosomatic NGS interpretation and can easily be implemented in other molecular biology laboratories. AI is becoming increasingly important in molecular biomedical analysis and can be very helpful in processing medical data. Neural networks show a good capacity in variant classification, and in the future, they may be useful in predicting more complex variants.

Association of tumor mutational burden with genomic alterations in Chinese urothelial carcinoma

The tumor mutational burden (TMB) calculated by whole-exome sequencing (WES) is a promising biomarker for the response to immune checkpoint inhibition (ICIs) in solid tumors. However, WES is not feasible in the routine clinical setting. In addition, the characteristics of the TMB in Chinese urothelial carcinoma (UC) are unclear. The aim of this study was to demonstrate the reliability of an Acornmed 808 panel and analyze the characteristics of the TMB in Chinese UC. An Acornmed 808 panel was designed and virtually validated using UC data from the cancer genome atlas (TCGA). Comprehensive analysis of sequencing and clinical data was performed to explore the characteristics of the TMB for 143 Chinese UC patients. Compared to the TMB calculated with random 808-, 500-, and 250-gene panels, the TMB calculated with the Acornmed 808 panel was closer to that calculated by WES. There were marked disparities in the mutational landscape and TMB between Chinese and TCGA UC data. The TMB was negatively associated with copy number variation (CNV). In contrast, the TMB was positive correlation with numbers of mutated DDR genes. Exposure to aristolochic acid signature was observed only in the TMB-high groups. The Acornmed 808 panel is a clinically practical method to assess the TMB. The TMB was associated with the DDR gene status and CNV counts and might be a biomarker for further stratification of UC patients. The study suggested that patients with high TMB may have a unique carcinogenic mechanism.

Validation of a Circulating Tumor DNA-Based Next-Generation Sequencing Assay in a Cohort of Patients with Solid tumors: A Proposed Solution for Decentralized Plasma Testing

BACKGROUND

Characterization of circulating tumor DNA (ctDNA) has been integrated into clinical practice. Although labs have standardized validation procedures to develop single locus tests, the efficacy of on-site plasma-based next-generation sequencing (NGS) assays still needs to be proved.

MATERIALS AND METHODS

In this retrospective study, we profiled DNA from matched tissue and plasma samples from 75 patients with cancer. We applied an NGS test that detects clinically relevant alterations in 33 genes and microsatellite instability (MSI) to analyze plasma cell-free DNA (cfDNA).

RESULTS

The concordance between alterations detected in both tissue and plasma samples was higher in patients with metastatic disease. The NGS test detected 77% of sequence alterations, amplifications, and fusions that were found in metastatic samples compared with 45% of those alterations found in the primary tumor samples (p = .00005). There was 87% agreement on MSI status between the NGS test and tumor tissue results. In three patients, MSI-high ctDNA correlated with response to immunotherapy. In addition, the NGS test revealed an FGFR2 amplification that was not detected in tumor tissue from a patient with metastatic gastric cancer, emphasizing the importance of profiling plasma samples in patients with advanced cancer.

CONCLUSION

Our validation experience of a plasma-based NGS assay advances current knowledge about translating cfDNA testing into clinical practice and supports the application of plasma assays in the management of oncology patients with metastatic disease. With an in-house method that minimizes the need for invasive procedures, on-site cfDNA testing supplements tissue biopsy to guide precision therapy and is entitled to become a routine practice.

IMPLICATIONS FOR PRACTICE

This study proposes a solution for decentralized liquid biopsy testing based on validation of a next-generation sequencing (NGS) test that detects four classes of genomic alterations in blood: sequence mutations (single nucleotide substitutions or insertions and deletions), fusions, amplifications, and microsatellite instability (MSI). Although there are reference labs that perform single-site comprehensive liquid biopsy testing, the targeted assay this study validated can be established locally in any lab with capacity to offer clinical molecular pathology assays. To the authors' knowledge, this is the first report that validates evaluating an on-site plasma-based NGS test that detects the MSI status along with common sequence alterations encountered in solid tumors.

Deep Learning Coordinate-Free Quantum Chemistry

Computing quantum chemical properties of small molecules and polymers can provide insights valuable into physicists, chemists, and biologists when designing new materials, catalysts, biological probes, and drugs. Deep learning can compute quantum chemical properties accurately in a fraction of time required by commonly used methods such as density functional theory. Most current approaches to deep learning in quantum chemistry begin with geometric information from experimentally derived molecular structures or pre-calculated atom coordinates. These approaches have many useful applications, but they can be costly in time and computational resources. In this study, we demonstrate that accurate quantum chemical computations can be performed without geometric information by operating in the coordinate-free domain using deep learning on graph encodings. Coordinate-free methods rely only on molecular graphs, the connectivity of atoms and bonds, without atom coordinates or bond distances. We also find that the choice of graph-encoding architecture substantially affects the performance of these methods. The structures of these graph-encoding architectures provide an opportunity to probe an important, outstanding question in quantum mechanics: what types of quantum chemical properties can be represented by local variable models? We find that Wave, a local variable model, accurately calculates the quantum chemical properties, while graph convolutional architectures require global variables. Furthermore, local variable Wave models outperform global variable graph convolution models on complex molecules with large, correlated systems.

Deep Modeling of Regulating Effects of Small Molecules on Longevity-Associated Genes

Aging is considered an inevitable process that causes deleterious effects in the functioning and appearance of cells, tissues, and organs. Recent emergence of large-scale gene expression datasets and significant advances in machine learning techniques have enabled drug repurposing efforts in promoting longevity. In this work, we further developed our previous approach-DeepCOP, a quantitative chemogenomic model that predicts gene regulating effects, and extended its application across multiple cell lines presented in LINCS to predict aging gene regulating effects induced by small molecules. As a result, a quantitative chemogenomic Deep Model was trained using gene ontology labels, molecular fingerprints, and cell line descriptors to predict gene expression responses to chemical perturbations. Other state-of-the-art machine learning approaches were also evaluated as benchmarks. Among those, the deep neural network (DNN) classifier has top-ranked known drugs with beneficial effects on aging genes, and some of these drugs were previously shown to promote longevity, illustrating the potential utility of this methodology. These results further demonstrate the capability of "hybrid" chemogenomic models, incorporating quantitative descriptors from biomarkers to capture cell specific drug-gene interactions. Such models can therefore be used for discovering drugs with desired gene regulatory effects associated with longevity.

Establishing guidelines to harmonize tumor mutational burden (TMB): in silico assessment of variation in TMB quantification across diagnostic platforms: phase I of the Friends of Cancer Research TMB Harmonization Project

Background

Tumor mutational burden (TMB), defined as the number of somatic mutations per megabase of interrogated genomic sequence, demonstrates predictive biomarker potential for the identification of patients with cancer most likely to respond to immune checkpoint inhibitors. TMB is optimally calculated by whole exome sequencing (WES), but next-generation sequencing targeted panels provide TMB estimates in a time-effective and cost-effective manner. However, differences in panel size and gene coverage, in addition to the underlying bioinformatics pipelines, are known drivers of variability in TMB estimates across laboratories. By directly comparing panel-based TMB estimates from participating laboratories, this study aims to characterize the theoretical variability of panel-based TMB estimates, and provides guidelines on TMB reporting, analytic validation requirements and reference standard alignment in order to maintain consistency of TMB estimation across platforms.

Methods

Eleven laboratories used WES data from The Cancer Genome Atlas Multi-Center Mutation calling in Multiple Cancers (MC3) samples and calculated TMB from the subset of the exome restricted to the genes covered by their targeted panel using their own bioinformatics pipeline (panel TMB). A reference TMB value was calculated from the entire exome using a uniform bioinformatics pipeline all members agreed on (WES TMB). Linear regression analyses were performed to investigate the relationship between WES and panel TMB for all 32 cancer types combined and separately. Variability in panel TMB values at various WES TMB values was also quantified using 95% prediction limits.

Results

Study results demonstrated that variability within and between panel TMB values increases as the WES TMB values increase. For each panel, prediction limits based on linear regression analyses that modeled panel TMB as a function of WES TMB were calculated and found to approximately capture the intended 95% of observed panel TMB values. Certain cancer types, such as uterine, bladder and colon cancers exhibited greater variability in panel TMB values, compared with lung and head and neck cancers.

Conclusions

Increasing uptake of TMB as a predictive biomarker in the clinic creates an urgent need to bring stakeholders together to agree on the harmonization of key aspects of panel-based TMB estimation, such as the standardization of TMB reporting, standardization of analytical validation studies and the alignment of panel-based TMB values with a reference standard. These harmonization efforts should improve consistency and reliability of panel TMB estimates and aid in clinical decision-making.

Development and Validation of StrataNGS, a Multiplex PCR, Semiconductor Sequencing-Based Comprehensive Genomic Profiling Test

Despite widespread use in targeted tumor testing, multiplex PCR/semiconductor (Ion Torrent) sequencing-based assessment of all comprehensive genomic profiling (CGP) variant classes has been limited. Herein, we describe the development and validation of StrataNGS, a 429-gene, multiplex PCR/semiconductor sequencing-based CGP laboratory-developed test performed on co-isolated DNA and RNA from formalin-fixed, paraffin-embedded tumor specimens with ≥2 mm² tumor surface area. Validation was performed in accordance with MolDX CGP validation guidelines using 1986 clinical formalin-fixed, paraffin-embedded samples and an in-house developed optimized bioinformatics pipeline. Across CGP variant classes, accuracy ranged from 0.945 for tumor mutational burden (TMB) status to >0.999 for mutations and gene fusions, positive predictive value ranged from 0.915 for TMB status to 1.00 for gene fusions, and reproducibility ranged from 0.998 for copy number alterations to 1.00 for splice variants and insertions/deletions. StrataNGS TMB estimates were highly correlated to those from whole exome- or FoundationOne CDx-determined TMB (Pearson r = 0.998 and 0.960, respectively); TMB reproducibility was 0.996 (concordance correlation coefficient). Limit of detection for all variant classes was <20% tumor content. Together, we demonstrate that multiplex PCR/semiconductor sequencing-based tumor tissue CGP is feasible using optimized bioinformatic approaches described herein.

Characterization of tumor mutation burden, PD-L1 and DNA repair genes to assess relationship to immune checkpoint inhibitors response in metastatic renal cell carcinoma

Background

Immune checkpoint inhibitors (ICIs) have expanded treatment options for metastatic renal cell carcinoma (mRCC); however, there are limited predictive biomarkers for response to ICIs in this indication, with programmed death-ligand 1 (PD-L1) status demonstrating little predictive utility in mRCC. While predictive of ICI response in other tumor types, the utility of tumor mutation burden (TMB) in mRCC is unclear. Here, we assess TMB, loss of antigen presentation genes and PD-L1 status correlated with outcomes to ICI treatment in mRCC.

Methods

Tumor samples from 34 patients with mRCC treated with ICI therapy at Duke Cancer Institute were retrospectively evaluated using Personal Genome Diagnostics elio tissue complete (RUO version), a tumor genomic profiling assay for somatic variants, TMB, microsatellite status and genomic status of antigen presentation genes. Tumor samples were also analyzed with the Dako 28-8 PD-L1 immunohistochemistry assay. Deidentified clinical information was extracted from the medical record, and tumor response was evaluated based on the Response Evaluation Criteria In Solid Tumors (RECIST) V.1.1 criteria.

Results

Patients were stratified by overall response following ICI therapy and designated as progressive disease (PD; n=18) or disease control groups (DC; n=16). TMB scores ranged from 0.36 to 12.24 mutations/Mb (mean 2.83 mutations/Mb) with no significant difference between the PD and DC groups (3.01 vs 2.63 mutations/Mb, respectively; p=0.7682). Interestingly, 33% of PD patients displayed loss of heterozygosity of major histocompatibility complex class I genes (LOH-MHC) vs 6% of DC patients. Nine of 34 samples were PD-L1-positive (4 in the PD group; 5 in the DC group), suggesting no correlation between PD-L1 expression and response to ICI therapy. Notably, the DC group displayed an enrichment of mutations in DNA repair genes (p=0.04), with 68.8% exhibiting at least one mutated homologous recombination repair (HRR)-related gene compared with only 38.9% of the PD group (p=0.03).

Conclusions

Overall, neither TMB nor PD-L1 correlated with ICI response and TMB was not significantly associated with PD-L1 expression. The higher incidence of LOH-MHC in PD group suggests that loss of antigen presentation may restrict response to ICIs. Separately, enrichment of HRR gene mutations in the DC group suggests potential utility in predicting ICI response and a potential therapeutic target, warranting future studies.

Real-World Performance of a Comprehensive Genomic Profiling Test Optimized for Small Tumor Samples

PURPOSE

Tissue-based comprehensive genomic profiling (CGP) is increasingly used for treatment selection in patients with advanced cancer; however, tissue availability may limit widespread implementation. Here, we established real-world CGP tissue availability and assessed CGP performance on consecutively received samples.

MATERIALS AND METHODS

We conducted a post hoc, nonprespecified analysis of 32,048 consecutive tumor tissue samples received for StrataNGS, a multiplex polymerase chain reaction (PCR)-based comprehensive genomic profiling (PCR-CGP) test, as part of an ongoing observational trial (NCT03061305). Sample characteristics and PCR-CGP performance were assessed across all tested samples, including exception samples not meeting minimum input quality control (QC) requirements (< 20% tumor content [TC], < 2 mm2 tumor surface area [TSA], DNA or RNA yield < 1 ng/µL, or specimen age > 5 years). Tests reporting ≥ 1 prioritized alteration or meeting TC and sequencing QC were considered successful. For prostate carcinoma and lung adenocarcinoma, tests reporting ≥ 1 actionable or informative alteration or meeting TC and sequencing QC were considered actionable.

RESULTS

Among 31,165 (97.2%) samples where PCR-CGP was attempted, 10.7% had < 20% TC and 59.2% were small (< 25 mm2 tumor surface area). Of 31,101 samples evaluable for input requirements, 8,089 (26.0%) were exceptions not meeting requirements. However, 94.2% of the 31,101 tested samples were successfully reported, including 80.5% of exception samples. Positive predictive value of PCR-CGP for ERBB2 amplification in exceptions and/or sequencing QC-failure breast cancer samples was 96.7%. Importantly, 84.0% of tested prostate carcinomas and 87.9% of lung adenocarcinomas yielded results informing treatment selection.

CONCLUSION

Most real-world tissue samples from patients with advanced cancer desiring CGP are limited, requiring optimized CGP approaches to produce meaningful results. An optimized PCR-CGP test, coupled with an inclusive exception testing policy, delivered reportable results for > 94% of samples, potentially expanding the proportion of CGP-testable patients and impact of biomarker-guided therapies.

Cloud Computing Enabled Big Multi-Omics Data Analytics

High-throughput experiments enable researchers to explore complex multifactorial diseases through large-scale analysis of omics data. Challenges for such high-dimensional data sets include storage, analyses, and sharing. Recent innovations in computational technologies and approaches, especially in cloud computing, offer a promising, low-cost, and highly flexible solution in the bioinformatics domain. Cloud computing is rapidly proving increasingly useful in molecular modeling, omics data analytics (eg, RNA sequencing, metabolomics, or proteomics data sets), and for the integration, analysis, and interpretation of phenotypic data. We review the adoption of advanced cloud-based and big data technologies for processing and analyzing omics data and provide insights into state-of-the-art cloud bioinformatics applications.

Next-Generation Sequencing Concordance Analysis of Comprehensive Solid Tumor Profiling between a Centralized Specialty Laboratory and the Decentralized Personal Genome Diagnostics, Inc., Elio Tissue Complete Kitted Solution

Genomic tumor profiling by next-generation sequencing (NGS) allows for large-scale tumor testing to inform targeted cancer therapies and immunotherapies, and to identify patients for clinical trials. These tests are often underutilized in patients with late-stage solid tumors and are typically performed in centralized specialty laboratories, thereby limiting access to these complex tests. Personal Genome Diagnostics Inc., elio tissue complete NGS solution is a comprehensive DNA-to-report kitted assay and bioinformatics solution. Comparison of 147 unique specimens from >20 tumor types was performed using the elio tissue complete solution and Foundation Medicine's FoundationOne test, which is of similar size and gene content. The analytical performance of all genomic variant types was evaluated. In general, the overall mutational profile is highly concordant between the two assays, with agreement in sequence variants reported between panels demonstrating >95% positive percentage agreement for single-nucleotide variants and insertions/deletions in clinically actionable genes. Both copy number alterations and gene translocations showed 80% to 83% positive percentage agreement, whereas tumor mutation burden and microsatellite status showed a high level of concordance across a range of mutation loads and tumor types. The Personal Genome Diagnostics Inc., elio tissue complete assay is comparable to the FoundationOne test and will allow more laboratories to offer a diagnostic NGS assay in house, which will ultimately reduce time to result and increase the number of patients receiving molecular genomic profiling and personalized treatment.

Second-Generation Digital Health Platforms: Placing the Patient at the Center and Focusing on Clinical Outcomes

Artificial intelligence (AI) digital health systems have drawn much attention over the last decade. However, their implementation into medical practice occurs at a much slower pace than expected. This paper reviews some of the achievements of first-generation AI systems, and the barriers facing their implementation into medical practice. The development of second-generation AI systems is discussed with a focus on overcoming some of these obstacles. Second-generation systems are aimed at focusing on a single subject and on improving patients' clinical outcomes. A personalized closed-loop system designed to improve end-organ function and the patient's response to chronic therapies is presented. The system introduces a platform which implements a personalized therapeutic regimen and introduces quantifiable individualized-variability patterns into its algorithm. The platform is designed to achieve a clinically meaningful endpoint by ensuring that chronic therapies will have sustainable effect while overcoming compensatory mechanisms associated with disease progression and drug resistance. Second-generation systems are expected to assist patients and providers in adopting and implementing of these systems into everyday care.

Uncovering the prognostic gene signatures for the improvement of risk stratification in cancers by using deep learning algorithm coupled with wavelet transform

Background

The aim of gene expression-based clinical modelling in tumorigenesis is not only to accurately predict the clinical endpoints, but also to reveal the genome characteristics for downstream analysis for the purpose of understanding the mechanisms of cancers. Most of the conventional machine learning methods involved a gene filtering step, in which tens of thousands of genes were firstly filtered based on the gene expression levels by a statistical method with an arbitrary cutoff. Although gene filtering procedure helps to reduce the feature dimension and avoid overfitting, there is a risk that some pathogenic genes important to the disease will be ignored.

Results

In this study, we proposed a novel deep learning approach by combining a convolutional neural network with stationary wavelet transform (SWT-CNN) for stratifying cancer patients and predicting their clinical outcomes without gene filtering based on tumor genomic profiles. The proposed SWT-CNN overperformed the state-of-art algorithms, including support vector machine (SVM) and logistic regression (LR), and produced comparable prediction performance to random forest (RF). Furthermore, for all the cancer types, we firstly proposed a method to weight the genes with the scores, which took advantage of the representative features in the hidden layer of convolutional neural network, and then selected the prognostic genes for the Cox proportional-hazards regression. The results showed that risk stratifications can be effectively improved by using the identified prognostic genes as feature, indicating that the representative features generated by SWT-CNN can well correlate the genes with prognostic risk in cancers and be helpful for selecting the prognostic gene signatures.

Conclusions

Our results indicated that gene expression-based SWT-CNN model can be an excellent tool for stratifying the prognostic risk for cancer patients. In addition, the representative features of SWT-CNN were validated to be useful for evaluating the importance of the genes in the risk stratification and can be further used to identify the prognostic gene signatures.

Tumor Mutational Burden as a Predictive Biomarker in Solid Tumors

Tumor mutational burden (TMB), defined as the number of somatic mutations per megabase of interrogated genomic sequence, varies across malignancies. Panel sequencing-based estimates of TMB have largely replaced whole-exome sequencing-derived TMB in the clinic. Retrospective evidence suggests that TMB can predict the efficacy of immune checkpoint inhibitors, and data from KEYNOTE-158 led to the recent FDA approval of pembrolizumab for the TMB-high tumor subgroup. Unmet needs include prospective validation of TMB cutoffs in relationship to tumor type and patient outcomes. Furthermore, standardization and harmonization of TMB measurement across test platforms are important to the successful implementation of TMB in clinical practice. SIGNIFICANCE: Evaluation of TMB as a predictive biomarker creates the need to harmonize panel-based TMB estimation and standardize its reporting. TMB can improve the predictive accuracy for immunotherapy outcomes, and has the potential to expand the candidate pool of patients for treatment with immune checkpoint inhibitors.

Site-Level Bioactivity of Small-Molecules from Deep-Learned Representations of Quantum Chemistry

Atom- or bond-level chemical properties of interest in medicinal chemistry, such as drug metabolism and electrophilic reactivity, are important to understand and predict across arbitrary new molecules. Deep learning can be used to map molecular structures to their chemical properties, but the data sets for these tasks are relatively small, which can limit accuracy and generalizability. To overcome this limitation, it would be preferable to model these properties on the basis of the underlying quantum chemical characteristics of small molecules. However, it is difficult to learn higher level chemical properties from lower level quantum calculations. To overcome this challenge, we pretrained deep learning models to compute quantum chemical properties and then reused the intermediate representations constructed by the pretrained network. Transfer learning, in this way, substantially outperformed models based on chemical graphs alone or quantum chemical properties alone. This result was robust, observable in five prediction tasks: identifying sites of epoxidation by metabolic enzymes and identifying sites of covalent reactivity with cyanide, glutathione, DNA and protein. We see that this approach may substantially improve the accuracy of deep learning models for specific chemical structures, such as aromatic systems.

Assembling and Validating Bioinformatic Pipelines for Next-Generation Sequencing Clinical Assays

CONTEXT.—

Clinical next-generation sequencing (NGS) is being rapidly adopted, but analysis and interpretation of large data sets prompt new challenges for a clinical laboratory setting. Clinical NGS results rely heavily on the bioinformatics pipeline for identifying genetic variation in complex samples. The choice of bioinformatics algorithms, genome assembly, and genetic annotation databases are important for determining genetic alterations associated with disease. The analysis methods are often tuned to the assay to maximize accuracy. Once a pipeline has been developed, it must be validated to determine accuracy and reproducibility for samples similar to real-world cases. In silico proficiency testing or institutional data exchange will ensure consistency among clinical laboratories.

OBJECTIVE.—

To provide molecular pathologists a step-by-step guide to bioinformatics analysis and validation design in order to navigate the regulatory and validation standards of implementing a bioinformatic pipeline as a part of a new clinical NGS assay.

DATA SOURCES.—

This guide uses published studies on genomic analysis, bioinformatics methods, and methods comparison studies to inform the reader on what resources, including open source software tools and databases, are available for genetic variant detection and interpretation.

CONCLUSIONS.—

This review covers 4 key concepts: (1) bioinformatic analysis design for detecting genetic variation, (2) the resources for assessing genetic effects, (3) analysis validation assessment experiments and data sets, including a diverse set of samples to mimic real-world challenges that assess accuracy and reproducibility, and (4) if concordance between clinical laboratories will be improved by proficiency testing designed to test bioinformatic pipelines.

Tumor mutational burden related classifier is predictive of response to PD-L1 blockade in locally advanced and metastatic urothelial carcinoma

BACKGROUND

Immunotherapy has made encouraging progress in the treatment of urothelial carcinoma, but only a small percentage of patients respond effectively to the immune checkpoint blockade (ICB). Our study aims to develop a classifier could effectively predict the response to ICB.

METHODS

Support vector machines recursive feature elimination (SVM-RFE) algorithm was used to feature selection, then compared nine common binary classification algorithms through machine learning, we selected the classifier with the highest prediction performance (LASSO logistics classifier). Ten-fold cross-validation was used to avoid the overfitting effect.

RESULTS

We developed a classifier on a urothelial carcinoma cohort treated with PD-L1 inhibitor Atzolizumab (IMvigor210 cohort, n = 272) and calculated a tumor mutational burden-related LASSO score (TLS) using the LASSO algorithm, which was significantly correlated with Tumor mutational burden (TMB) and neoantigen burden. We validated the efficacy of TLS in predicting prognosis and immunotherapy benefit in internal (IMvigor210) and external validation set (TCGA-BLCA, n = 406), respectively.

CONCLUSIONS

After in-depth analysis, we provide a new idea for stratified treatment of such patients, that is, patients with high TLS should use ICB and also may benefit from hormone therapy, while patients with low TLS respond poorly to ICB and maybe benefit from targeting TGFβ.

SMuRF: portable and accurate ensemble prediction of somatic mutations

Summary

Somatic Mutation calling method using a Random Forest (SMuRF) integrates predictions and auxiliary features from multiple somatic mutation callers using a supervised machine learning approach. SMuRF is trained on community-curated matched tumor and normal whole genome sequencing data. SMuRF predicts both SNVs and indels with high accuracy in genome or exome-level sequencing data. Furthermore, the method is robust across multiple tested cancer types and predicts low allele frequency variants with high accuracy. In contrast to existing ensemble-based somatic mutation calling approaches, SMuRF works out-of-the-box and is orders of magnitudes faster.

Availability and implementation

The method is implemented in R and available at https://github.com/skandlab/SMuRF. SMuRF operates as an add-on to the community-developed bcbio-nextgen somatic variant calling pipeline.

Supplementary information

Supplementary data are available at Bioinformatics online.

Precision Health for Chagas Disease: Integrating Parasite and Host Factors to Predict Outcome of Infection and Response to Therapy

Chagas disease, caused by the infection with the protozoan parasite Trypanosoma cruzi, is clinically manifested in approximately one-third of infected people by inflammatory heart disease (cardiomyopathy) and, to a minor degree, gastrointestinal tract disorders (megaesophagus or megacolon). Chagas disease is a zoonosis transmitted among animals and people through the contact with triatomine bugs, which are found in much of the western hemisphere, including most countries of North, Central and South America, between parallels 45° north (Minneapolis, USA) and south (Chubut Province, Argentina). Despite much research on drug discovery for T. cruzi, there remain only two related agents in widespread use. Likewise, treatment is not always indicated due to the serious side effects of these drugs. On the other hand, the epidemiology and pathogenesis of Chagas disease are both highly complex, and much is known about both. However, it is still impossible to predict what will happen in an individual person infected with T. cruzi, because of the highly variability of parasite virulence and human susceptibility to infection, with no definitive molecular predictors of outcome from either side of the host-parasite equation. In this Minireview we briefly discuss the current state of T. cruzi infection and prognosis and look forward to the day when it will be possible to employ precision health to predict disease outcome and determine whether and when treatment of infection may be necessary.

A highly accurate platform for clone-specific mutation discovery enables the study of active mutational processes

Bulk whole genome sequencing (WGS) enables the analysis of tumor evolution but, because of depth limitations, can only identify old mutational events. The discovery of current mutational processes for predicting the tumor’s evolutionary trajectory requires dense sequencing of individual clones or single cells. Such studies, however, are inherently problematic because of the discovery of excessive false positive (FP) mutations when sequencing picogram quantities of DNA. Data pooling to increase the confidence in the discovered mutations, moves the discovery back in the past to a common ancestor. Here we report a robust WGS and analysis pipeline (DigiPico/MutLX) that virtually eliminates all F results while retaining an excellent proportion of true positives. Using our method, we identified, for the first time, a hyper-mutation (kataegis) event in a group of ∼30 cancer cells from a recurrent ovarian carcinoma. This was unidentifiable from the bulk WGS data. Overall, we propose DigiPico/MutLX method as a powerful framework for the identification of clone-specific variants at an unprecedented accuracy.

Exploring the Consistency of the Quality Scores with Machine Learning for Next-Generation Sequencing Experiments

Background

Next-generation sequencing enables massively parallel processing, allowing lower cost than the other sequencing technologies. In the subsequent analysis with the NGS data, one of the major concerns is the reliability of variant calls. Although researchers can utilize raw quality scores of variant calling, they are forced to start the further analysis without any preevaluation of the quality scores.

Method

We presented a machine learning approach for estimating quality scores of variant calls derived from BWA+GATK. We analyzed correlations between the quality score and these annotations, specifying informative annotations which were used as features to predict variant quality scores. To test the predictive models, we simulated 24 paired-end Illumina sequencing reads with 30x coverage base. Also, twenty-four human genome sequencing reads resulting from Illumina paired-end sequencing with at least 30x coverage were secured from the Sequence Read Archive.

Results

Using BWA+GATK, VCFs were derived from simulated and real sequencing reads. We observed that the prediction models learned by RFR outperformed other algorithms in both simulated and real data. The quality scores of variant calls were highly predictable from informative features of GATK Annotation Modules in the simulated human genome VCF data (R2: 96.7%, 94.4%, and 89.8% for RFR, MLR, and NNR, respectively). The robustness of the proposed data-driven models was consistently maintained in the real human genome VCF data (R2: 97.8% and 96.5% for RFR and MLR, respectively).