ECP 2022 Abstract Book

Virchows Archiv (2022) 481 (Suppl 1):S1–S364 13 using Oncomine Comprehensive (Thermofisher) NGS. BRG1 expression was evaluated by immunohistochemistry and cor- related with SMARCA4. Medical records were retrospectively reviewed for clinicopathologic, molecular characteristics and treatment outcomes. Survival analyses were performed using Kaplan-Meier. Results: We detected SMARCA4 genomic alterations in 11.6% (n=34/292) of NSCLCs. Truncating missense mutations com- prised 22 cases, nonsense 9, frameshift 3 and INDEL 2. 44.1% of SMARCA4-mutant NSCLCs (n=15/34) showed loss of expres- sion of BRG1, most (80%) of which had truncating SMARCA4 mutations. Overall, 92% (n=12/13) of evaluated NSCLCs with nonsense, frameshift or INDEL mutations lacked BRG1 expres- sion. Deficient BRG1 expression was detected predominantly in solid adenocarcinomas (G3) and NSCLC-NOS with co-occurring mutations in KRAS (n=8; 3 G12C), TP53 (n=4), STK11 (n=4) and MAP2K2 (n=3). Deficiency on BRG1 was associated with lower disease-free survival (log rank p=0.049) and a tendency to associate with lower overall survival (p=0.2) in comparison with non-BRG1-deficient. Conclusion: BRG1 deficiency is enriched in NSCLCs with trun- cating SMARCA4 mutations. Clinical outcomes are poor in this molecular subgroup, highlighting the importance of developing novel strategies to target unique vulnerabilities associated with the BRG1-deficient state. PS-15-020 Validation of RNAseq as first approach: results of the real- world study at the Lorraine Cancer Institute V. Haddad*, A. Harle, M. Husson, J. Dardare, A. Lambert, M. Betz, A. Witz, P. Gilson, J. Merlin *Invitae Corporation, San Francisco, CA, USA, France Background & objectives: Next-generation sequencing has become the standard for tumour sequencing, but most laboratories use sequen- tial workflows. We optimized our workflow by using RNA-based FusionPlex® assay as the primary approach and assessed the ability of this assay to detect fusions and mutations. Methods: A total of 43 formalin-fixed paraffin embedded (FFPE) tissues from patients with various cancer types (colorectal, lung, sarcoma, and pancreatic cancers) were included in this study. After RNA and DNA extraction, all samples were assessed using the FusionPlex® assay for gene fusions, SNVs/indels and custom STS, for the detection of SNVs /indels and sequenced using Illu- mina platform or Idylla assay. Results: Repeatability of the assay was assessed using a triplicate within the same run, and reproducibility by analysing 3 different samples in two different runs. Limit of detection was evaluated using RNA input from 20 to 250ng. Among the 43 analysed samples, 35 SNVs or indels were previously identified using DNAseq or Idylla. The total concordance for indels and SNVs for ALK, BRAF, EGFR, KRAS, RET and ROS1 genes was 97.1% (Sensitivity=0.97, Specificity=1.0). Only one mutation G12V of KRAS with a VAF of 3.3% has not been found using RNAseq. Repeatability and reproducibility of the assay were both 100%. Finally, 20ng of RNA were found sufficient as a minimal input. Conclusion: Analytical concordance, repeatability, reproduc- ibility, and a robust limit of detection was documented in this assessment of the RNAseq approach. This data supports the RNAseq first approach, using FusionPlex® research assay (Invi- tae), and a decision has been made to adopt this workflow in our laboratory. PS-15-021 Proof of concepts for automated extraction of mutation status from narrative pathology reports: KRAS G12C mutations and NTRK fusions in non-squamous non-small cell lung cancer V. de Jager*, B. Cajiao Garcia, B. Koopman, V.A. Stangenberger, R.L. Rijnders, R.M. de Valk, E. Schuuring, S. Willems, L.C. van Kempen *University of Groningen, University Medical Center Groningen, Department of Pathology and Medical Biology, Groningen, The Netherlands Background & objectives: Manual extraction of molecular diagnostic results from narrative pathology reports is labour intensive, causing delay in analysis of real-world data. The aim of this study is to develop an automated method for data extraction from narrative pathology reports. Methods: For automated selection of reports indicating the presence of the KRAS p.(G12C) mutation, a 5-fold cross validation Random Forest model was performed using TF-IDF transformed report text. For analysis of NTRK fusions, an SVC model was used with a supplementary word-association model. Previous manually extracted results from PALGA reports were used as a gold-standard. Results: Pathology reports deposited in the Dutch nationwide pathology databank (PALGA) of all NSCLC patients in 2015 were manually curated (n=2,427) for reporting of the KRAS p.(G12C). The algorithm achieved high sensitivity (98.7%) and specificity (100%) to identify this mutation in narrative text. In addition, pathology reports mentioning TRK or NTRK between 2017 and 2020 were manually curated (n=7,457) for reporting of TRK protein expression and/or NTRK fusions. A first validation cohort demonstrated an acceptable sensitivity (88.0%) and speci- ficity (95.0%) of the algorithm. TRK expression was generally not described as a dichotomous result, leading to large variation in reporting. All false-negative cases were ambiguous reporting of TRK immunohistochemistry results. Conclusion: The algorithm that was developed for identification of pathology reports mentioning KRAS p.(G12C) demonstrated high accuracy. In contrast to KRAS mutations, the prevalence of TRK expression or NTRK fusion is low, which hampers the training of the algorithm with only a low absolute number of unambiguous TRK/ NTRK positive cases. The observed variety in reporting of TRK results requires further optimization of the TRK/ NTRK algo- rithm. Nevertheless, automation-supported analysis of pathology reports is a promising tool for rapid assessment of real-world data. PS-15-022 KRAS-G12C mutation in non-small cell lung cancer: preva- lence and prognostic significance V. Cristóbal Redondo*, V. Pedrero Castillo, D. Parra Trujillo, C. Alenda González, F.I. Aranda López *General Hospital of Alicante. ISABIAL, Spain Background & objectives: KRAS-G12C variant, which has recently been proved to be druggable, is the most frequent KRAS mutation, accounting for the 13% of non-small cell lung cancer (NSCLC). We aim to determine its prevalence, concurrent mutations, histopathologic profile and prognostic value. Methods: We used Next-Generation Sequencing with the Oncomine Comprehensive panel to identify KRAS mutation sta- tus and other genetic alterations in 400 patients from our institu- tion. Pathological findings were retrospectively reviewed using previous biopsy slides. Clinical data was extracted from medical records, including age, gender, smoking history, treatment, stage, S146