Automated Analysis of 1p/19q Status by FISH in Oligodendroglial Tumors: Rationale and Proposal of an Algorithm

Efficient diagnosis of IDH-mutant gliomas: 1p/19qNET assesses 1p/19q codeletion status using weakly-supervised learning

Accurate identification of molecular alterations in gliomas is crucial for their diagnosis and treatment. Although, fluorescence in situ hybridization (FISH) allows for the observation of diverse and heterogeneous alterations, it is inherently time-consuming and challenging due to the limitations of the molecular method. Here, we report the development of 1p/19qNET, an advanced deep-learning network designed to predict fold change values of 1p and 19q chromosomes and classify isocitrate dehydrogenase (IDH)-mutant gliomas from whole-slide images. We trained 1p/19qNET on next-generation sequencing data from a discovery set (DS) of 288 patients and utilized a weakly-supervised approach with slide-level labels to reduce bias and workload. We then performed validation on an independent validation set (IVS) comprising 385 samples from The Cancer Genome Atlas, a comprehensive cancer genomics resource. 1p/19qNET outperformed traditional FISH, achieving R2 values of 0.589 and 0.547 for the 1p and 19q arms, respectively. As an IDH-mutant glioma classifier, 1p/19qNET attained AUCs of 0.930 and 0.837 in the DS and IVS, respectively. The weakly-supervised nature of 1p/19qNET provides explainable heatmaps for the results. This study demonstrates the successful use of deep learning for precise determination of 1p/19q codeletion status and classification of IDH-mutant gliomas as astrocytoma or oligodendroglioma. 1p/19qNET offers comparable results to FISH and provides informative spatial information. This approach has broader applications in tumor classification.

Diagnostic accuracy of 1p/19q codeletion tests in oligodendroglioma: A comprehensive meta-analysis based on a Cochrane systematic review

Codeletion of chromosomal arms 1p and 19q, in conjunction with a mutation in the isocitrate dehydrogenase 1 or 2 gene, is the molecular diagnostic criterion for oligodendroglioma, IDH mutant and 1p/19q codeleted. 1p/19q codeletion is a diagnostic marker and allows prognostication and prediction of the best drug response within IDH-mutant tumours. We performed a Cochrane review and simple economic analysis to establish the most sensitive, specific and cost-effective techniques for determining 1p/19q codeletion status. Fluorescent in situ hybridisation (FISH) and polymerase chain reaction (PCR)-based loss of heterozygosity (LOH) test methods were considered as reference standard. Most techniques (FISH, chromogenic in situ hybridisation [CISH], PCR, real-time PCR, multiplex ligation-dependent probe amplification [MLPA], single nucleotide polymorphism [SNP] array, comparative genomic hybridisation [CGH], array CGH, next-generation sequencing [NGS], mass spectrometry and NanoString) showed good sensitivity (few false negatives) for detection of 1p/19q codeletions in glioma, irrespective of whether FISH or PCR-based LOH was used as the reference standard. Both NGS and SNP array had a high specificity (fewer false positives) for 1p/19q codeletion when considered against FISH as the reference standard. Our findings suggest that G banding is not a suitable test for 1p/19q analysis. Within these limits, considering cost per diagnosis and using FISH as a reference, MLPA was marginally more cost-effective than other tests, although these economic analyses were limited by the range of available parameters, time horizon and data from multiple healthcare organisations.

Validation of genomic and transcriptomic models of homologous recombination deficiency in a real-world pan-cancer cohort

Background

With the introduction of DNA-damaging therapies into standard of care cancer treatment, there is a growing need for predictive diagnostics assessing homologous recombination deficiency (HRD) status across tumor types. Following the strong clinical evidence for the utility of DNA-sequencing-based HRD testing in ovarian cancer, and growing evidence in breast cancer, we present analytical validation of the Tempus HRD-DNA test. We further developed, validated, and explored the Tempus HRD-RNA model, which uses gene expression data from 16,750 RNA-seq samples to predict HRD status from formalin-fixed paraffin-embedded tumor samples across numerous cancer types.

Methods

Genomic and transcriptomic profiling was performed using next-generation sequencing from Tempus xT, Tempus xO, Tempus xE, Tempus RS, and Tempus RS.v2 assays on 48,843 samples. Samples were labeled based on their BRCA1, BRCA2 and selected Homologous Recombination Repair pathway gene (CDK12, PALB2, RAD51B, RAD51C, RAD51D) mutational status to train and validate HRD-DNA, a genome-wide loss-of-heterozygosity biomarker, and HRD-RNA, a logistic regression model trained on gene expression.

Results

In a sample of 2058 breast and 1216 ovarian tumors, BRCA status was predicted by HRD-DNA with F1-scores of 0.98 and 0.96, respectively. Across an independent set of 1363 samples across solid tumor types, the HRD-RNA model was predictive of BRCA status in prostate, pancreatic, and non-small cell lung cancer, with F1-scores of 0.88, 0.69, and 0.62, respectively.

Conclusions

We predict HRD-positive patients across many cancer types and believe both HRD models may generalize to other mechanisms of HRD outside of BRCA loss. HRD-RNA complements DNA-based HRD detection methods, especially for indications with low prevalence of BRCA alterations.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-022-09669-z.

Diagnostic test accuracy and cost-effectiveness of tests for codeletion of chromosomal arms 1p and 19q in people with glioma

BACKGROUND

Complete deletion of both the short arm of chromosome 1 (1p) and the long arm of chromosome 19 (19q), known as 1p/19q codeletion, is a mutation that can occur in gliomas. It occurs in a type of glioma known as oligodendroglioma and its higher grade counterpart known as anaplastic oligodendroglioma. Detection of 1p/19q codeletion in gliomas is important because, together with another mutation in an enzyme known as isocitrate dehydrogenase, it is needed to make the diagnosis of an oligodendroglioma. Presence of 1p/19q codeletion also informs patient prognosis and prediction of the best drug treatment. The main two tests in use are fluorescent in situ hybridisation (FISH) and polymerase chain reaction (PCR)-based loss of heterozygosity (LOH) assays (also known as PCR-based short tandem repeat or microsatellite analysis). Many other tests are available. None of the tests is perfect, although PCR-based LOH is expected to have very high sensitivity.

OBJECTIVES

To estimate the sensitivity and specificity and cost-effectiveness of different deoxyribonucleic acid (DNA)-based techniques for determining 1p/19q codeletion status in glioma.

SEARCH METHODS

We searched MEDLINE, Embase and BIOSIS up to July 2019. There were no restrictions based on language or date of publication. We sought economic evaluation studies from the results of this search and using the National Health Service Economic Evaluation Database.

SELECTION CRITERIA

We included cross-sectional studies in adults with glioma or any subtype of glioma, presenting raw data or cross-tabulations of two or more DNA-based tests for 1p/19q codeletion. We also sought economic evaluations of these tests.

DATA COLLECTION AND ANALYSIS

We followed procedures outlined in the Cochrane Handbook for Diagnostic Test Accuracy Reviews. Two review authors independently screened titles/abstracts/full texts, performed data extraction, and undertook applicability and risk of bias assessments using QUADAS-2. Meta-analyses used the hierarchical summary ROC model to estimate and compare test accuracy. We used FISH and PCR-based LOH as alternate reference standards to examine how tests compared with those in common use, and conducted a latent class analysis comparing FISH and PCR-based LOH. We constructed an economic model to evaluate cost-effectiveness.

MAIN RESULTS

We included 53 studies examining: PCR-based LOH, FISH, single nucleotide polymorphism (SNP) array, next-generation sequencing (NGS), comparative genomic hybridisation (CGH), array comparative genomic hybridisation (aCGH), multiplex-ligation-dependent probe amplification (MLPA), real-time PCR, chromogenic in situ hybridisation (CISH), mass spectrometry (MS), restriction fragment length polymorphism (RFLP) analysis, G-banding, methylation array and NanoString. Risk of bias was low for only one study; most gave us concerns about how patients were selected or about missing data. We had applicability concerns about many of the studies because only patients with specific subtypes of glioma were included. 1520 participants contributed to analyses using FISH as the reference, 1304 participants to analyses involving PCR-based LOH as the reference and 262 participants to analyses of comparisons between methods from studies not including FISH or PCR-based LOH. Most evidence was available for comparison of FISH with PCR-based LOH (15 studies, 915 participants): PCR-based LOH detected 94% of FISH-determined codeletions (95% credible interval (CrI) 83% to 98%) and FISH detected 91% of codeletions determined by PCR-based LOH (CrI 78% to 97%). Of tumours determined not to have a deletion by FISH, 94% (CrI 87% to 98%) had a deletion detected by PCR-based LOH, and of those determined not to have a deletion by PCR-based LOH, 96% (CrI 90% to 99%) had a deletion detected by FISH. The latent class analysis suggested that PCR-based LOH may be slightly more accurate than FISH. Most other techniques appeared to have high sensitivity (i.e. produced few false-negative results) for detection of 1p/19q codeletion when either FISH or PCR-based LOH was considered as the reference standard, although there was limited evidence. There was some indication of differences in specificity (false-positive rate) with some techniques. Both NGS and SNP array had high specificity when considered against FISH as the reference standard (NGS: 6 studies, 243 participants; SNP: 6 studies, 111 participants), although we rated certainty in the evidence as low or very low. NGS and SNP array also had high specificity when PCR-based LOH was considered the reference standard, although with much more uncertainty as these results were based on fewer studies (just one study with 49 participants for NGS and two studies with 33 participants for SNP array). G-banding had low sensitivity and specificity when PCR-based LOH was the reference standard. Although MS had very high sensitivity and specificity when both FISH and PCR-based LOH were considered the reference standard, these results were based on only one study with a small number of participants. Real-time PCR also showed high specificity with FISH as a reference standard, although there were only two studies including 40 participants. We found no relevant economic evaluations. Our economic model using FISH as the reference standard suggested that the resource-optimising test depends on which measure of diagnostic accuracy is most important. With FISH as the reference standard, MLPA is likely to be cost-effective if society was willing to pay GBP 1000 or less for a true positive detected. However, as the value placed on a true positive increased, CISH was most cost-effective. Findings differed when the outcome measure changed to either true negative detected or correct diagnosis. When PCR-based LOH was used as the reference standard, MLPA was likely to be cost-effective for all measures of diagnostic accuracy at lower threshold values for willingness to pay. However, as the threshold values increased, none of the tests were clearly more likely to be considered cost-effective.

AUTHORS' CONCLUSIONS

In our review, most techniques (except G-banding) appeared to have good sensitivity (few false negatives) for detection of 1p/19q codeletions in glioma against both FISH and PCR-based LOH as a reference standard. However, we judged the certainty of the evidence low or very low for all the tests. There are possible differences in specificity, with both NGS and SNP array having high specificity (fewer false positives) for 1p/19q codeletion when considered against FISH as the reference standard. The economic analysis should be interpreted with caution due to the small number of studies.

Shallow whole-genome sequencing: a useful, easy to apply molecular technique for CNA detection on FFPE tumor tissue-a glioma-driven study

Copy number alterations (CNAs) have increasingly become part of the diagnostic algorithm of glial tumors. Alterations such as homozygous deletion of CDKN2A/B, 7 +/ 10 - chromosome copy number changes or EGFR amplification are predictive of a poor prognosis. The codeletion of chromosome arms 1p and 19q, typically associated with oligodendroglioma, implies a more favorable prognosis. Detection of this codeletion by the current diagnostic standard, being fluorescence in situ hybridization (FISH), is sometimes however subject to technical and interpretation problems. In this study, we evaluated CNA detection by shallow whole-genome sequencing (sWGS) as an inexpensive, complementary molecular technique. A cohort of 36 glioma tissue samples, enriched with "difficult" and "ambiguous" cases, was analyzed by sWGS. sWGS results were compared with FISH assays of chromosomes 1p and 19q. In addition, CNAs relevant to glioblastoma diagnosis were explored. In 4/36 samples, EGFR (7p11.2) amplifications and homozygous loss of CDKN2A/B were identified by sWGS. Six out of 8 IDH-wild-type glioblastomas demonstrated a prognostic chromosome 7/chromosome 10 signature. In 11/36 samples, local interstitial and terminal 1p/19q alterations were detected by sWGS, implying that FISH's targeted nature might promote false arm-level extrapolations. In this cohort, differences in overall survival between patients with and without codeletion were better pronounced by the sequencing-based distinction (likelihood ratio of 7.48) in comparison to FISH groupings (likelihood ratio of 0.97 at diagnosis and 1.79 ± 0.62 at reobservation), suggesting sWGS is more accurate than FISH. We recognized adverse effects of tissue block age on FISH signals. In addition, we show how sWGS reveals relevant aberrations beyond the 1p/19q state, such as EGFR amplification, combined gain of chromosome 7 and loss of chromosome 10, and homozygous loss of CDKN2A/B. The findings presented by this study might stimulate implementation of sWGS as a complementary, easy to apply technique for copy number detection.

Impact of 9p deletion and p16, Cyclin D1, and Myc hyperexpression on the outcome of anaplastic oligodendrogliomas

Objective

To study the presence of 9p deletion and p16, cyclin D1 and Myc expression and their respective diagnostic and prognostic interest in oligodendrogliomas.

Methods

We analyzed a retrospective series of 40 consecutive anaplastic oligodendrogliomas (OIII) from a single institution and compared them to a control series of 10 low grade oligodendrogliomas (OII). Automated FISH analysis of chromosome 9p status and immunohistochemistry for p16, cyclin D1 and Myc was performed for all cases and correlated with clinical and histological data, event free survival (EFS) and overall survival (OS).

Results

Chromosome 9p deletion was observed in 55% of OIII (22/40) but not in OII. Deletion was highly correlated to EFS (median = 29 versus 53 months, p<0.0001) and OS (median = 48 versus 83 months, p<0.0001) in both the total cohort and the OIII population. In 9p non-deleted oligodendrogliomas, p16 hyperexpression correlated with a shorter OS (p = 0.02 in OII and p = 0.0001 in OIII) whereas lack of p16 expression was correlated to a shorter EFS and OS in 9p deleted OIII (p = 0.001 and p = 0.0002 respectively). Expression of Cyclin D1 was significantly higher in OIII (median expression 45% versus 14% for OII, p = 0.0006) and was correlated with MIB-1 expression (p<0.0001), vascular proliferation (p = 0.002), tumor necrosis (p = 0.04) and a shorter EFS in the total cohort (p = 0.05). Hyperexpression of Myc was correlated to grade (median expression 27% in OII versus 35% in OIII, p = 0.03), and to a shorter EFS in 9p non-deleted OIII (p = 0.01).

Conclusion

Chromosome 9p deletion identifies a subset of OIII with significantly worse prognosis. The combination of 9p status and p16 expression level identifies two distinct OIII populations with divergent prognosis. Hyperexpression of Bcl1 and Myc appears highly linked to anaplasia but the prognostic value is unclear and should be investigated further.

Contribution of 1p, 19q, 9p and 10q Automated Analysis by FISH to the Diagnosis and Prognosis of Oligodendroglial Tumors According to WHO 2016 Guidelines

Objective

To study the feasibility and the diagnostic and prognostic interest of automated analysis of 1p, 19q, 9p and 10q status by FISH technique in oligodendroglial tumors.

Methods

We analyzed a retrospective series of 33 consecutive gliomas with oligodendroglial histology (originally diagnosed as 24 oligodendrogliomas and 9 oligoastrocytomas). For all cases, automated FISH analysis of 1p, 19q, 9p and 10q status were performed and compared to clinical and histological data, ATRX, IDH1R132H and alpha-internexin status (studied by immunohistochemistry) and overall survival (OS). Manual analysis of 9p and 10q status were also performed and compared to automated analysis to verify the concordance of the two methods.

Results

The 33 gliomas were reclassified into 13 low-grade oligodendrogliomas (OII), 10 anaplastic oligodendrogliomas (OIII), 3 diffuse astrocytomas (AII), 3 anaplastic astrocytomas (AIII) and 4 glioblastomas (GBM) according to the WHO 2016 histological criteria. The 1p and/or 19q imbalanced status were restricted to astrocytomas with no correlation to their grade or their OS. Chromosome 9p deletion was restricted to OIII (70%) and GBM (100%) and was correlated with a shorter OS in the total cohort (p = 0.0007), the oligodendroglioma cohort (p = 0.03) and the astrocytoma cohort (p = 0.001). Concordance between 9p manual and automated analysis was satisfactory (81%, κ = 0.69). Chromosome 10q deletion was restricted to GBMs (50%) and was correlated with a poor OS in both the total cohort (p = 0.003) and the astrocytoma (AS) cohort (p = 0.04). Concordance between manual and automated analysis was satisfactory (79%, κ = 0.62).

Conclusion

Automated analysis of 1p, 19q, 9p and 10q status by FISH is a reliable technique which allows for refined classification of oligodendroglial tumors. 1p and/or 19q imbalanced status is evidence of astrocytic differentiation. 9p deletion is found in high grade oligodendrogliomas and astrocytomas with a poor OS. 10q is related to GBM status and a poor OS.