Reliable Pan-Cancer Microsatellite Instability Assessment by Using Targeted Next-Generation Sequencing Data
A.Z. and J.F.H. contributed equally to this work.
Microsatellite instability (MSI)/mismatch repair (MMR) status is increasingly important in the management of patients with cancer to predict response to immune checkpoint inhibitors. We determined MSI status from large-panel clinical targeted next-generation sequencing (NGS) data across various solid cancer types.
The MSI statuses of 12,288 advanced solid cancers consecutively sequenced with Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets clinical NGS assay were inferred by using MSIsensor, a program that reports the percentage of unstable microsatellites as a score. Cutoff score determination and sensitivity/specificity were based on MSI polymerase chain reaction (PCR) and MMR immunohistochemistry.
By using an MSIsensor score ≥ 10 to define MSI high (MSI-H), 83 (8%) of 996 colorectal cancers (CRCs) and 42 (16%) of 260 uterine endometrioid cancers (UECs) were MSI-H. Validation against MSI PCR and/or MMR immunohistochemistry performed for 138 (24 MSI-H, 114 microsatellite stable [MSS]) CRCs, and 40 (15 MSI-H, 25 MSS) UECs showed a concordance of 99.4%. MSIsensor also identified 68 MSI-H/MMR-deficient (MMR-D) non-CRC/UECs. Of 9,591 non-CRC/UEC tumors with MSS MSIsensor status, 456 (4.8%) had slightly elevated scores (≥ 3 and < 10) of which 96.6% with available material were confirmed to be MSS by MSI PCR. MSI-H was also detected and confirmed in three non-CRC/UECs with low exonic mutation burden (< 20). MSIsensor correctly scored all 15 polymerase ε ultra-mutated cancers as negative for MSI.
MSI status can be reliably inferred by MSIsensor from large-panel targeted NGS data. Concurrent MSI testing by NGS is resource efficient, is potentially more sensitive for MMR-D than MSI PCR, and allows identification of MSI-H across various cancers not typically screened, as highlighted by the finding that 35% (68 of 193) of all MSI-H tumors were non-CRC/UEC.
Microsatellites are short, tandemly repeated DNA sequences of 1 to 6 bases scattered throughout the human genome. These sites are prone to DNA replication errors as a result of DNA polymerase slippage, which is effectively corrected through the mismatch repair (MMR) system. Deficiencies in MMR result in increased variation at genomic loci with mononucleotide repeats. Microsatellite instability (MSI) testing often is used to screen MMR protein status, and MSI polymerase chain reaction (PCR) and MMR immunohistochemistry (IHC) testing are particularly important for the clinical management of both colorectal cancer (CRC) and uterine endometrioid cancer (UEC). The National Comprehensive Cancer Network recommends MSI PCR/MMR IHC testing for all patients with CRC1,2 and for patients with UEC at risk for Lynch syndrome.1 MSI/MMR status has implications for prognosis,3 screening for Lynch syndrome, and response to fluorouracil3 and immune checkpoint inhibitor therapy.4 Recently, the Food and Drug Administration granted pembrolizumab accelerated approval as the first drug approved for any solid tumor with a specific genetic feature (MSI-high [MSI-H] status) on the basis of new data that confirm its activity across 12 different cancer types, with complete responses observed in 21% of patients.5
Until now, the gold standard for assessment of MSI, a reliable screen for functional MMR status, has been concurrent analysis of patient tumor and normal DNA for five mononucleotide microsatellite loci with PCR. The gold standard for detecting MMR protein expression status has been IHC for MLH1, MSH2, PMS2, and MSH6 expression. In recent years, reports have shown that next-generation sequencing (NGS) facilitates identification of patients with deficiencies in the MMR pathway by comparing sequencing reads around microsatellite regions in the tumor and the matched normal or by counting mutations identified in exons. Hause et al6 identified MSI/MMR across a wide spectrum of tumor types surveyed by The Cancer Genome Atlas but with limited validation data available in only a subset of CRC/UECs and stomach cancers. Although MSI PCR and MMR IHC are not routinely performed in all cancer types, many patients with solid malignancies of all types at our center undergo molecular testing for somatic alterations with the NGS clinical assay Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT),7 and we have recently reported an analysis of the data on the first 10,000 patients studied.8 In the current study, we investigated the sensitivity and specificity of assessing MSI by using NGS data across all solid cancers tested prospectively and validated this NGS-based method of MSI assessment with MSI PCR and MMR IHC.
After approval by our local institutional review board, data from 12,288 patients who underwent molecular testing with MSK-IMPACT between January 1, 2014, and December 31, 2016, were analyzed. MSK-IMPACT is an NGS assay that uses tumor and matched normal DNA to identify somatic mutations, structural variants, and copy number alterations in all coding regions and select introns of 341 (version 1), 410 (version 2), or 468 (version 3) cancer-related genes.7 Tumor purity (TP) was estimated with a combination of median variant allele frequency of mutations identified in each sample and microscopic analysis of hematoxylin and eosin–stained specimens.
MSIsensor interrogates the aligned sequencing data for available microsatellite regions with sufficient coverage in a tumor/normal pair where it identifies deletion length variation. χ2 test is used to identify the significantly varied loci, and the percentage of unstable loci, after multiple testing correction is performed on the P values, is reported as an MSIsensor score; additional details have been previously published.9 This score was used to infer MSI/MMR status from NGS data in the current study. Cross-validation with MSI PCR and MMR IHC was used to establish a cutoff criterion for MSI-H on the basis of the MSIsensor score of tumor specimens against a matched normal blood sample. Validation specimens also were subjected to unmatched analysis in which a pool of 10 normal samples was used as a comparator. We presumed that the use of 10 normal samples would dilute out any inherent unstable loci a normal sample might have and that the pooled normal would be a relatively stable comparator.
MMR IHC data for CRC and UEC were retrospectively collected for patients who underwent clinical IHC testing, and MMR IHC was performed for non-CRC/UECs with discordant MSI PCR and MSIsensor scores. All non-CRC/UECs with MSIsensor scores ≥ 3 with available material underwent MSI PCR testing.
Analysis of IHC expression of MMR proteins was performed by using a clinically validated standard streptavidin-biotin-peroxidase procedure. Primary monoclonal antibodies used were MLH1 (clone G168-728, diluted 1:250; BD Biosciences Pharmingen, San Diego, CA), MSH2 (clone FE11, diluted 1:50; Oncogene Research Products, Cambridge, MA), MSH6 (clone GRBP.P1/2.D4, diluted 1:200; AbD Serotec, Raleigh, NC), and PMS2 (clone A16-4, diluted 1:200; BD Biosciences Pharmingen). Non-neoplastic colonic mucosa and colorectal tumors known to be deficient of MLH1, MSH2, MSH6, and PMS2 were used as external positive and negative controls, respectively. Retained expression of each protein was defined by nuclear IHC reactivity of tumor cells, whereas loss of expression for each protein was defined by the complete absence of nuclear IHC reactivity of tumor cells. Tumors were MMR proficient if all four proteins were expressed (retained) by IHC and MMR deficient (MMR-D) if any of the four proteins was not expressed (lost) by IHC.
Analysis of MSI status by PCR was performed by using a commercially available kit (MSI Analysis System, version 1.2, catalog #MD1641; Promega, Madison, WI) that has been clinically validated. The PCR assay assesses the spectrum of the number of nucleotides in five mononucleotide microsatellite loci, including NR-21, BAT-25, MONO-27, NR-24, and BAT-26, in both tumor and normal DNA. A shift of ≥ 3 base pairs in the tumor DNA compared with the match normal constitutes instability at one locus. Instability at two or more of the five microsatellite loci defines MSI-H status. For the purposes of this study, all tumors with fewer than two unstable loci were interpreted as microsatellite stable (MSS).
Proper detection of MSI-H/MMR-D status from NGS data by using MSIsensor depends on whether the tumor is analyzed against its matched normal sample, because analysis with an unmatched sample results in significant inflation of the MSIsensor scores (P < .001; Appendix Fig A1). Alignment files generated from the MSK-IMPACT pipeline in binary alignment map format were analyzed by using MSIsensor on all matched tumor-normal pairs and then separately on all unmatched tumor samples to generate the score. Cutoff determination was performed by using the train function from the caret package in R version 3.2. Cross-validation was done by partitioning the data into five groups.
Dilution experiments were performed by diluting formalin-fixed paraffin-embedded tumor DNA with a high MSIsensor score (validated with MSI PCR) with successive amounts of normal DNA from matched formalin-fixed paraffin-embedded tissue. MSIsensor was run at the following successive dilutions: 100% tumor DNA, 50% tumor DNA and 50% normal DNA, 25% tumor DNA and 75% normal DNA, 12% tumor DNA and 88% normal DNA, and 6% tumor DNA and 94% normal DNA. We chose an MSI-H CRC specimen with original TP of approximately 70% and MSIsensor score of 36.7 (undiluted). At successive dilutions of 50% (TP, 35%) and 25% (TP, 17.5%), the MSIsensor score remained relatively consistent (34.5 and 26.9, respectively). Additional dilutions of 12% (TP, 8.4%) and 6% (TP, 4.2%) resulted in a larger MSIsensor score difference (17.1 and 8.2, respectively; Appendix Table A1). We used a total TP threshold of ≥ 25% for this application of MSIsensor.
Tumor mutation burden (TMB) was calculated by dividing the total number of reported mutations by the genomic area where mutations were reported for each sample. To further investigate the MSS samples with higher mutation rates, we identified samples with TMB > 11.85 nonsynonymous mutations/megabase (calculated as median TMB + 2 × interquartile range TMB). Contributions of different mutation signatures were identified for each sample according to the distribution of the six substitution classes (C>A, C>G, C>T, T>A, T>C, T>G) and the bases immediately 5′ and 3′ of the mutated base, which produced 96 possible mutation subtypes. These mutations were resampled 1,000 times and then subjected to decomposition analysis in which the Kullback-Leibler divergence is minimized between the sample signature and the approximation built up from 30 signatures such that each signature is assigned a weight that corresponds to the percentage of mutations explained by each given signature. A sample was determined to have a dominant signature if > 40% of observed mutations were attributable to that signature.10
Time from procedure date to last follow-up was used for survival analysis among patients whose biopsy date and sequencing date were within 1 year of each other. MSS and MSI-H group survival differences were compared by using a Cox proportional hazards regression model adjusted for age. Kaplan-Meier curves were used to visualize the differences between groups. The survival package in R version 3.2 was used for the survival analysis.
One hundred thirty-eight CRC and 40 UEC samples with matched normals were selected for orthogonal MSI/MMR status by PCR/IHC and used as the training data set for MSIsensor score threshold determination. MSIsensor scores ranged from 0 to 47.7 and 0 to 43.7 for CRC and UEC, respectively. We used the classification and regression tree method to identify a cutoff MSIsensor score that would separate MSS from MSI-H tumors identified by IHC and/or PCR. We used five-fold cross-validation and found that an MSIsensor score of 9.3 to 10 could delineate MSS from MSI-H with 99.4% accuracy. For ease of interpretation, we used 10 as our cutoff criterion for interpreting MSIsensor scores (Figs 1A and 1B; Table 1). This set included four MSH6-deficient CRCs and one MSH6-equivocal UEC, all with concordant MSIsensor scores (> 10).
We then evaluated the entire study cohort of 13,091 cancer samples from 12,288 patients with 66 principle cancer types sequenced between January 2014 and December 2016 against the MSIsensor cutoff score. From 10,900 patients, 11,553 samples had sufficient coverage of 200× and TP of ≥ 25% [as described in Analytic (Technical) Sensitivity] and matched normal samples for analysis. MSIsensor scores ranged from 0.0 to 48.5 (mean, 1.2; median, 0.4; Appendix Fig A2).
Two hundred four samples from 193 patients (1.8% of cohort; 20 tumor types) displayed an MSI-H phenotype by MSIsensor (Fig 2). Among the patients without CRC/UEC, bladder cancer (11 [3.1%] of 355), esophagogastric carcinoma (seven [2.5%] of 282), and prostate cancer (12 [1.7%] of 722) had MSI-H incidence. Of 68 patients without CRC/UEC with MSIsensor scores ≥ 10, 49 had available material for MSI PCR/MMR IHC. Forty-six were concordant (PCR, n = 39; IHC, n = 7), whereas three were MSI low (MSI-L) by PCR but were MMR-D by IHC (Figs 1C-1G), including a carcinoma of unknown primary with 16 mutations, a squamous cell carcinoma with 65 mutations, and a prostate carcinoma with 44 mutations. Overall, we observed 100% concordance of MSI-H non-CRC/UECs between MSIsensor and IHC and/or PCR. Of note, use of the MSIsensor on targeted panel NGS data had higher clinical sensitivity as a screen for MMR-D than MSI PCR.
MSIsensor was originally used to assess MSI in UEC from The Cancer Genome Atlas whole-exome data and validated at a cutoff score of 3.5 from an average of 10,000 evaluable microsatellites per sample.9 MSK-IMPACT covers less of the genome with higher coverage, with approximately 1,000 evaluable microsatellites per sample. This number has increased with the addition of genes to the MSK-IMPACT panel (Appendix Fig A3). With consideration of the different cutoffs between the present and previous studies, we also evaluated the 456 (4.8%) of 9,591 MSK-IMPACT non-CRC/UECs with slightly elevated MSIsensor scores (≥ 3 and < 10). MSI PCR was performed on 58 of these: 47 (81%) were MSS, nine were MSI-L (15.5%), and only two (3.5%) were discordant (MSI-H) by PCR (one melanoma with two unstable loci and one esophageal adenocarcinoma with three unstable loci; Fig 3). Of note, three samples (a germ cell tumor with two mutations, a cancer of unknown primary with 16 mutations, and a uterine sarcoma with 19 mutations) that were MSI-H by both MSIsensor and PCR had < 20 exonic mutations in MSK-IMPACT, a cutoff shown to be highly predictive of MSI-H in CRC samples.9
Because most clinical laboratories that offer NGS testing use a pooled unrelated normal rather than a matched normal, we investigated whether and at what cutoff MSIsensor could be used for unmatched tumors. Our set of orthogonally validated (through PCR/IHC) tumors was analyzed with a pooled rather than an unmatched normal. A cutoff MSIsensor score of 25.6 differentiated between MSS and MSI-H/MMR-D status with 96.1% sensitivity and 98.5% specificity (Appendix Fig A4).
TMB has been used previously to infer MSI status in a CRC cohort.11 However, other mutagenic signatures (smoking, UV exposure or BRCA1/2 deficiency, polymerase ε [POLE]) are also known to result in increased TMB. We looked for signatures of mutational processes in samples with a sufficient number of mutations (as described in the Methods). Of 825 MSS (MSIsensor score < 10) samples with elevated TMB (> 11.85 nonsynonymous mutations/megabase), 56.5% had signatures other than MMR-D (Fig 2). The most common were UV exposure, APOBEC deficiency, and smoking (22.5%, 20%, and 10.7% of samples, respectively).
Fifteen of these highly mutated tumors (total mutation range, 35 to 456) had POLE signatures, all with known hotspot mutations in POLE12 (Table 2). Only one of these had an MSI-H result by MSIsensor, whereas the rest had scores < 10. The MSI-H tumor was a UEC that was MSH6 deficient. MMR IHC was available for another 10 of these POLE mutations, and all were concordant (MMR proficient).
MSI-H status has been shown to be associated with better prognosis. Our analysis of survival status for MSI-H compared with MSS in cancer types where we identified at least 10 patients with MSI-H status showed no significant difference after adjusting for age (Fig 4). However, within CRC, we identified a strong trend of better survival for patients with MSI-H status. Given that this cohort is a heterogeneous group of patients with complex treatment histories, additional examination of the relationship between MSI/MMR status and survival is warranted.
We show that MSI status can be reliably inferred from NGS data across many solid tumor types by using MSIsensor. The sensitivity and specificity for MSI-H in CRC and UEC were 100% and 99.3%, respectively, whereas sensitivity and specificity (of patients with slightly elevated MSIsensor scores) across other tumor types were 96.6% and 100%, respectively.
As a result of the growing number of patients with CRC and UEC who undergo both MSI/MMR and NGS testing, increasing interest exists in MSI testing with NGS data6,9,13-17 because the National Comprehensive Cancer Network has recently recommended universal MSI/MMR testing for CRC, immune checkpoint inhibitors that are effective in CRC have gained Food and Drug Administration approval for any advanced MSI-H solid cancers that have not responded to previous therapies,5 and tissue is limited to small biopsy specimens in metastatic settings. The current validation study is one of the largest to date and shows that MSI status can be inferred from NGS data on a clinical, pan-tumor basis.
One hundred ninety-three of 10,900 total patients with 20 tumor types had MSI-H status (MSIsensor score ≥ 10), with 16.2%, 15.6%, and 8.3% of small bowel carcinomas, UECs, and CRCs identified as MSI-H (Appendix Tables A2 and A3). The rate of MSI-H is lower in our tumor set than a recent large investigation of MSI-H prevalence across 18 tumor types6 possibly because our data set is mostly restricted to patients with advanced (metastatic) cancer. Although the bias in our data set may underestimate the total number of MSI-H tumors across all patients with cancer, it gives a more accurate estimate of the proportion of patients with advanced solid cancers who may be eligible for immune checkpoint inhibition on the basis of MSI-H status.
Although mutation rate has been used to infer MSI status in CRC,10,14 various other hypermutation signatures (POLE, UV, smoking, temozolomide) also are associated with an increased mutation rate in other tumor types. Very few (three of 10,900) patients with MMR-D/MSI-H status do not display a high mutation burden. MSIsensor allowed higher sensitivity and specificity than mutation burden for MSI-H status compared with mutation burden for rare MSI-H tumors with low mutation rates and MSS tumors with hypermutation.
Furthermore, we demonstrate that MSIsensor may be more sensitive for the detection of MMR-D than MSI PCR in rare tumors that are MSI-L or MSS yet MMR-D on IHC. Although MSI PCR only tests five or seven loci, NGS-based methods scan hundreds to thousands of available loci, which allows for a more-thorough assessment. Although research has shown that MSI PCR is less sensitive for MSH6-deficient tumors, none of the three MMR-D cancers with high MSIsensor scores and MSS/MSI-L PCR results were MSH6 deficient.17 All four MSH6-deficient CRCs and one MSH6-equivocal UEC were MSI-H on MSIsensor assessment, which suggests that MSIsensor performs well in MSH6-deficient cancers.
Conversely, two MSI-H tumors were missed by MSIsensor. Because so few (4.8%) non-CRC/UECs had slightly elevated MSIsensor scores that did not reach the MSI-H cutoff, review of this subset and performance of MSI PCR/MMR IHC when TP/coverage is borderline may be prudent.
In conclusion, NGS-based MSI testing is both highly sensitive and highly specific when TP and coverage are > 25% and 200×, respectively. Programs such as MSIsensor allow for differentiation between MSI-H and other hypermutation signatures and identification of rare MSI-H/MMR-D cancers with lower mutation rates.
Conception and design: Sumit Middha, Liying Zhang, Gowtham Jayakumaran, Deborah F. Delair, Jinru Shia, David S. Klimstra, Marc Ladanyi, Ahmet Zehir, Jaclyn F. Hechtman
Administrative support: David S. Klimstra
Provision of study material or patients: Liying Zhang, Justyna Sadowska, Jinru Shia
Collection and assembly of data: Sumit Middha, Liying Zhang, Khedoudja Nafa, Gowtham Jayakumaran, Justyna Sadowska, Deborah F. Delair, Ahmet Zehir, Jaclyn F. Hechtman
Data analysis and interpretation: Sumit Middha, Liying Zhang, Khedoudja Nafa, Gowtham Jayakumaran, Donna Wong, Hyunjae R. Kim, Michael F. Berger, Deborah F. Delair, Jinru Shia, Zsofia Stadler, Marc Ladanyi, Ahmet Zehir, Jaclyn F. Hechtman
Manuscript writing: All authors
Final approval of manuscript: All authors
Accountable for all aspects of the work: All authors
The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or po.ascopubs.org/site/ifc.
No relationship to disclose
No relationship to disclose
No relationship to disclose
No relationship to disclose
No relationship to disclose
No relationship to disclose
No relationship to disclose
Consulting or Advisory Role: Cancer Genetics, Sequenom
No relationship to disclose
No relationship to disclose
Consulting or Advisory Role: Allergan (I), Genentech (I), Roche (I), Regeneron Pharmaceuticals (I), Optos (I), Adverum Biotechnologies (I)
Stock and Other Ownership Interests: PAIGE.AI
Consulting or Advisory Role: Wren Laboratories, Ipsen
Honoraria: Merck (I)
Consulting or Advisory Role: National Comprehensive Cancer Network/Boehringer Ingelheim Afatinib Targeted Therapy Advisory Committee, National Comprehensive Cancer Network/AstraZeneca Tagrisso RFP Advisory Committee
Research Funding: Loxo (Inst)
No relationship to disclose
Consulting or Advisory Role: Navigant Consulting
|1.||Benson AB III, Venook AP, Cederquist L, et al: Colon cancer, version 1.2017, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 15:370-398, 2017 Crossref, Medline, Google Scholar|
|2.||Hechtman JF, Middha S, Stadler ZK, et al: Universal screening for microsatellite instability in colorectal cancer in the clinical genomics era: new recommendations, methods, and considerations. Fam Cancer 10.1007/s10689-017-9993-x [epub ahead of print on April 12, 2017] Google Scholar|
|3.||Ribic CM, Sargent DJ, Moore MJ, et al: Tumor microsatellite-instability status as a predictor of benefit from fluorouracil-based adjuvant chemotherapy for colon cancer. New Engl J Med 349:247-257, 2003 Google Scholar|
|4.||Le DT, Uram JN, Wang H, et al: PD-1 blockade in tumors with mismatch-repair deficiency. N Engl J Med 372:2509-2520, 2015 Google Scholar|
|5.||Le DT, Durham JN, Smith KN, et al: Mismatch repair deficiency predicts response of solid tumors to PD-1 blockade. Science 357:409-413, 2017 Crossref, Medline, Google Scholar|
|6.||Hause RJ, Pritchard CC, Shendure J, et al: Classification and characterization of microsatellite instability across 18 cancer types. Nat Med 22:1342-1350, 2016 Crossref, Medline, Google Scholar|
|7.||Cheng DT, Mitchell TN, Zehir A, et al: Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): A hybridization capture-based next-generation sequencing clinical assay for solid tumor molecular oncology. J Mol Diagn 17:251-264, 2015 Crossref, Medline, Google Scholar|
|8.||Zehir A, Benayed R, Shah RH, et al: Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients. Nat Med 23:703-713, 2017 Google Scholar|
|9.||Niu B, Ye K, Zhang QY, et al: MSIsensor: Microsatellite instability detection using paired tumor-normal sequence data. Bioinformatics 30:1015-1016, 2014 Google Scholar|
|10.||GitHub: MSKCC/mutational signatures. https://github.com/mskcc/mutation-signatures Google Scholar|
|11.||Stadler ZK, Battaglin F, Middha S, et al: Reliable detection of mismatch repair deficiency in colorectal cancers using mutational load in next-generation sequencing panels. J Clin Oncol 34:2141-2147, 2016 Google Scholar|
|12.||Henninger EE, Pursell ZF: DNA polymerase ε and its roles in genome stability. IUBMB Life 66:339-351, 2014 Crossref, Medline, Google Scholar|
|13.||Salipante SJ, Scroggins SM, Hampel HL, et al: Microsatellite instability detection by next generation sequencing. Clin Chem 60:1192-1199, 2014 Crossref, Medline, Google Scholar|
|14.||Hempelmann JA, Scroggins SM, Pritchard CC, et al: MSIplus for integrated colorectal cancer molecular testing by next-generation sequencing. J Mol Diagn 17:705-714, 2015 Crossref, Medline, Google Scholar|
|15.||Nowak JA, Yurgelun MB, Bruce JL, et al: Detection of mismatch repair deficiency and microsatellite instability in colorectal adenocarcinoma by targeted next-generation sequencing. J Mol Diagn 19:84-91, 2017 Google Scholar|
|16.||Huang MN, McPherson JR, Cutcutache I, et al: MSIseq: Software for assessing microsatellite instability from catalogs of somatic mutations. Sci Rep 5:13321, 2015 Google Scholar|
|17.||Kautto EA, Bonneville R, Miya J, et al: Performance evaluation for rapid detection of pan-cancer microsatellite instability with MANTIS. Oncotarget 8:7452-7463, 2017 Google Scholar|
|18.||Shia J. Immunohistochemistry versus microsatellite instability testing for screening colorectal cancer patients at risk for hereditary nonpolyposis colorectal cancer syndrome. Part I. The utility of immunohistochemistry. J Mol Diagn 10:293-300, 2008 Google Scholar|