Skip to main content

RNA sequencing validation of the Complexity INdex in SARComas prognostic signature.

Lesluyes T, Pérot G, Largeau MR, Brulard C, Lagarde P, Dapremont V, Lucchesi C, Neuville A, Terrier P, Vince-Ranchère D, Mendez-Lago M, Gut M, Gut I, Coindre JM, Chibon F

Publication: Eur J Cancer (2016) • Vol 57 • Page 104-11PMID: 26916546View on PubMed

Purpose

This paper compared microarray and next generation RNA sequencing (RNAseq) gene expression profiles of frozen sarcoma specimens and also compared RNAseq data obtained in matched FFPE and frozen specimens.

Conclusion

The correlation in expression between microarray and RNAseq data was modest (r=0.6232), but was strong when considering only the Complexity INdex in SARComas (CINSARC) genes (r= 0.8402). Application of CINSARC centroids based on microarray date was prognostic in another cohort of sarcomas analyzed by microarray and by RNAseq and for specimens in the Cancer Genome Atlas (TCGA). As expected, when the number of reads decreased the correlation in CINSARC classification declined and prognostic value was not significant when the data was adjusted to reflect only 2 million reads.

Correlations in expression between FFPE and frozen specimens were weak to modest for 22 specimens (r=0.2-0.6) and modest to strong for 19 specimens (r=0.6-0.9) with significantly higher correlations observed for specimens with better FFPE RNA quality. However, FFPE specimens had significantly less ACTB and FPKM compared to frozen specimens, regardless of FFPE RNA quality. The quality of the RNA was not significantly affected by block or RNA storage; tumor type, size, necrosis, or location; or patient treatment.

Studies (2)

Purpose

The purpose of this study was to compare gene expression profiles obtained from microarray and RNAseq data of frozen sarcoma specimens and to determine the effect of number of reads. This study included microarray data from 217 frozen sarcomas and matched microarray and RNAseq data from a second cohort of 95 frozen sarcomas. The tissue location of the sarcomas was unspecified but they contained at least 60% tumor cells and less than 5% necrosis. RNA was extracted using Trizol with RNeasy Mini Kit. Specimens were analyzed using Affymetrix Human Genome U133 Plus 2.0 microarrays and RNAseq libraries were constructed using TruSeq RNA Sample Prep Kit v2 and an Illumina HiSeq2000.

Summary of Findings

Of the 19,179 genes investigated, 12,671 were expressed at greater than 1.0 expected fragments per kilobase of transcript per million fragments sequenced (FPKM), 15,097 at greater than 0.1 FPKM, and 16,480 at greater than 0.01 FPKM. Overall, the correlation in expression between microarray and RNAseq data was modest (r=0.6232), but was strong when considering only the Complexity INdex in SARComas (CINSARC) genes (r= 0.8402). Application of metastasis-free survival centroids that were computed based on the microarray data from 217 frozen specimens was prognostic in another cohort of 95 frozen sarcomas analyzed by microarray (P=0.028) and by RNAseq (P=0.013). Similarly, centroids that were calculated based on the RNAseq data from 95 frozen sarcomas were prognostic for the cohort of 217 specimens analyzed by microarray (P=0.000097).  Further, use of the prognostic signature based on centroids had prognostic value for a further 246 sarcoma specimens in the Cancer Genome Atlas (TCGA, P=0.0041). When the centroid of transcripts of CINSARC genes that were the best predictor of metastatic outcome were applied to TCGA, a significant prognostic value was observed (P=0.0079) with 79% of specimens classified as high risk of metastasis. The correlation of CINSARC classification between the unmodified RNAseq data and that obtained by computationally reducing the number of reads declined as the read number was decreased (R=.9892, R= 0.9817 and R=0.9492 for 20, 10 and 2 million reads, respectively). However, even after computationally decreasing the reads, the CINSARC classification remained the same for 95%, 94%, and 80% of tumors at 20, 10, and 2 million reads; but the prognostic value was only significant for 20 and 10 million reads (P=0.016 and P=0.032, respectively).

  • Biospecimens

    TissueMuscle (Skeletal)
    TissueAdipose
    TissueBone
  • Preservation Types

    Frozen
  • Diagnoses

    NeoplasticSarcoma
  • Platforms

    RNANext generation sequencing
    RNADNA microarray
  • Pre-analytical Factors

    ClassificationPre-Analytical FactorValue(s)
    Next generation sequencing SpecificData handling
    Adjusted to reflect 10 million reads
    Adjusted to reflect 2 million reads
    Adjusted to reflect 20 million reads
    Unadjusted (>70 million reads)
    Next generation sequencing SpecificTechnology platform
    Microarray
    RNAseq