Predictive Gene Expression Signatures Research Articles

12423 Pre-treatment Blood Transcriptome Accurately Predicts Growth Response To Daily And Weekly GH Treatment In Children Born Small For Gestational Age

Abstract Disclosure: T. Garner: Grant Recipient; Self; Novo Nordisk. P. Murray: Grant Recipient; Self; Novo Nordisk. M. Hojby: Employee; Self; Novo Nordisk. Stock Owner; Self; Novo Nordisk. R. Ard: Employee; Self; Novo Nordisk. Stock Owner; Self; Novo Nordisk. P.E. Clayton: Grant Recipient; Self; Novo Nordisk. A. Stevens: Grant Recipient; Self; Novo Nordisk. Background: Predicting GH therapy response from gene expression has the potential to improve clinical management of short stature. Treatment in children born small for gestational age (SGA) requires daily GH injections. Although no long-acting GHs (LAGHs) are currently approved for SGA, results from REAL5 (randomised, multinational, open-label, controlled, phase 2 trial; NCT03878446) indicate that the LAGH somapacitan has an efficacy, safety, and tolerability profile similar to daily GH (1). Here, we investigate growth response prediction based on the baseline blood transcriptome with and without clinical variables in daily GH- or somapacitan-treated children in REAL5. Methods: REAL5 tested three somapacitan doses (0.16, 0.20, & 0.24 mg/kg/week; n=12, n=13, & n=12, respectively) versus two daily GH doses (0.035 & 0.067 mg/kg/day; n=12 & n=13, respectively). GH response was explored using four metrics calculated over 52-weeks: height velocity (HV; cm/year), HV standard deviation score (SDS), change in height SDS, and change in IGF-I SDS. Due to small sample sizes, dose groups were combined. Participants were split into tertiles to define fastest/slowest responders to each treatment which were compared to remaining tertiles. Differentially expressed genes (DEGs) were defined to distinguish one tertile from the remaining two. Random forest (RF) models were generated using the top 100 DEGs as well as in combination with a set of clinical variables. Area under the curve (AUC) describes RF prediction accuracy: for somapacitan sufficient samples were available for both training and validation while for daily growth hormone samples were only sufficient for training generating out of bag (OOB) AUCs, accompanied by error rates (ER). Accuracies were calculated for both fast & slow responders across all response metrics in each treatment. Results: We demonstrate excellent predictive ability of the transcriptome to identify SGA children who respond slowest in HV following treatment with either daily GH or somapacitan (OOB AUC (ER)= 1.0 (5.0%) & 0.99 (6.9%), respectively; validation AUC=0.812 for somapacitan), while prediction of other response measures was highly variable, particularly in validation (AUCs=0.56-1.0). The clinical variables combined with the transcriptomic data resulted in reduced accuracies and increased error rates (OOBs AUC(ER)=0.93(25%) [daily] & 0.90(19%) [somapacitan]). Conclusions: In the first use of transcriptomics for growth response prediction in children born SGA we show good performance for those treated with both daily and weekly GH. However, when the transcriptome is combined with clinical data performance is reduced, indicating the importance of gene expression signatures for GH-response prediction. These findings may be promising for personalizing dose in GH-treated SGA patients. (1) Juul et al. EJE 2023:188,19-30. Presentation: 6/3/2024

Abstract LB092: Novel tumor gene expression signatures predictive of outcome in large B cell lymphoma treated with CAR T cell therapy (axicabtagene ciloleucel)

Abstract We leveraged ZUMA-7, the first and largest, phase III randomized CAR-T cell clinical study of relapsed/refractory large B cell lymphoma (LBCL) in the 2nd line setting, to discover tumor gene expression (GE) signatures associated with outcome to anti-CD19 CAR-T cell therapy, Axicabtagene Ciloleucel (axi-cel). We performed GE profiling on pre-treatment LBCL tumor samples from 134 axi-cel-treated (safety dataset) patients using IO-360 Nanostring panel (769 genes). Multivariate analysis with penalized Cox models was conducted to identify genes whose expression (transcripts) associated with event-free survival (EFS), progression-free survival (PFS), and duration of response (DOR), leading to identification of two predictive GE signatures. For technical validation, the predictive value of these signatures was reproduced in a subset of ZUMA-7 samples profiled by RNA-seq (n=124). High levels (> median) of a favorable 6-transcript GE signature (6-GES) composed of CD19, CD45RA, CCL22, KLRK1, SOX11, and SIGLEC5 positively correlated with EFS (HR: 0.27, 95% CI: 0.16–0.44; P=1.82 × 10−8) and PFS (HR: 0.27, 95% CI: 0.16–0.46; P=1.35 × 10−7) in patients treated with axi-cel. The 6-GES may capture lymphomas with abundant adhesion molecules, a relatively low inflammation, and abundant expression of the targeted antigen (CD19). Conversely, high levels of an unfavorable 17-transcript GE signature (17-GES; CD45RO, BCL2, IL-18R1, TNFSF4 [OX40L], KLRB1 [CD161], KIR3DL2, ITGB8, DUSP5, GPC4, PSMB5, RPS6KB1, SERPINA9, NBN, GLUD1, ESR1, ARID1A, and SLC16A1) negatively correlated with Axi-cel EFS (HR: 6.19, 95% CI: 3.60–10.65; P=1.51 × 10−13) and PFS (HR: 7.58, 95% CI: 4.16–13.81; P=2.70 × 10−14). The 17-GES signature is consistent with a high level of immune infiltration and inflammation paralleled by the activation of immune-escape mechanisms, such as the upregulation of anti-apoptotic genes. Notably, the 17-GES was elevated at progression after axi-cel treatment (n=18). These signatures did not associate with outcome to 2nd-line SOC from ZUMA-7 (n = 122 for nanostring and n = 125 for ZUMA-7 RNAseq datasets, respectively), nor with outcome to 1st -line R-CHOP from two online datasets, indicating their predictive rather than prognostic value. In conclusion, our transcriptomic analysis of ZUMA-7 dataset identified novel GE signatures predictive of outcome with axi-cel. These GE signatures could support risk-stratification of LBCL patients. Citation Format: Yuan Tian, Justin Budka, Frederick L. Locke, Jason R. Westin, Christina To, Shilpa Shahani, Davide Bedognetti, Rhine R. Shen, Jorge Andrade, Simone Filosto. Novel tumor gene expression signatures predictive of outcome in large B cell lymphoma treated with CAR T cell therapy (axicabtagene ciloleucel) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB092.

Abstract 6826: Comparative analysis of a predictive transcriptomic model and functional gene expression signatures (FGES) for tertiary lymphoid structure (TLS) detection in lung adenocarcinoma (LUAD)

Abstract Background: TLS are recognized as significant biomarkers for patient prognosis and therapy selection. In current clinical practice, TLS are often detected by manual assessment of H&E slides, which has been associated with reproducibility issues. Here, we developed a transcriptomic model using TLS density to stratify LUAD patients into TLS-high and -low groups and compared it with 5 published TLS FGES to assess the predictive value for patient outcomes. Design: A regression model to predict TLS density (units(u)/mm²) was created using RNA expression data and histological mapping of 415 TCGA-LUAD samples. Features were chosen based on differential gene expression analysis of sample groups with low (<Q1, 0.26 u/mm²) and high (>Q3, 2.04 u/mm²) TLS density. The model was trained on 269 samples and independently validated on 146 TCGA and 9 internal samples. The RNA model and five previously reported TLS FGES were compared by Wilcoxon paired test to Spearman’s ρ and F1-scores from bootstrapped validation set. The resulting predictions stratified immunotherapy-treated samples from SU2C-MARK (n=58) and GSE182328 (n=28) into TLS-high and TLS-low groups by Q2 threshold for TLS density (0.78 u/mm²) for further univariate (Kaplan-Meier) and multivariate (Cox model, adjusted for age and sex) analyses for overall survival (OS) with log-rank tests. P-values were adjusted with Bonferroni correction, and 95% confidence intervals (CIs) were provided. Results: Statistically significant upregulated genes in samples with high TLS density included FDCSP, MS4A1, and CXCL13 (logFC ≥ 0.95, FDR<0.01). The transcriptomic model showed confident superiority in correlation with TLS density compared to the best FGES in TCGA samples (ρ=0.53±0.03 vs 0.41±0.03, p<0.001). The TLS model also outperformed the highest performing FGES when stratifying TCGA samples into TLS-high and TLS-low groups (F1=0.73±0.02 vs 0.69±0.02, p<0.005). In the internal cohort, the model demonstrated robust performance (F1=0.76±0.05, ρ=0.81±0.04) and outperformed all FGES except one (FGES F1=0.78±0.05, ρ=0.86±0.05). The TLS-high group defined by the TLS RNA model had a significantly better prognosis on immunotherapy cohorts (p=0.01; 1 year OS 49% for TLS-low vs 70% for TLS-high), whereas no discernible difference in OS was identified for the 5 TLS FGES. The TLS model had a log hazard ratio (logHR) of -3.11 (CI:-5.31,-0.91; p<0.05, Harrell's C-index=0.64), indicative of a stronger association with OS than of the TLS FGES (logHR=-2.11; CI:-4.1,-0.08; p=ns; C-index=0.59). Conclusions: Comparative analysis of the proposed transcriptomic model for TLS detection in LUAD samples demonstrated improved performance metrics and prognostic value compared to previously reported TLS FGES, affirming its potential for guiding therapeutic decision-making. Citation Format: Nadezhda Lukashevich, Daniil Dymov, Aleksandr Sarachakov, Sofya Kust, Anna Love, Ivan Valiev, Dmitry Ivchenkov, Alexander Bagaev, Viktor Svekolkin, Nikita Kotlov, Vladimir Kushnarev. Comparative analysis of a predictive transcriptomic model and functional gene expression signatures (FGES) for tertiary lymphoid structure (TLS) detection in lung adenocarcinoma (LUAD) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6826.

Inflammation-related prognostic markers in resected hepatocellular carcinoma.

Hepatocellular carcinoma is usually detected late and therapeutic options are unsatisfactory. Despite marked progress in patient care, HCC remains among the deadliest cancers world-wide. While surgical resection remains a key option for early-stage HCC, the 5-year survival rates after surgical resection are limited. One reason for limited outcomes is the lack of reliable prognostic biomarkers to predict HCC recurrence. HCC prognosis has been shown to correlate with different systemic and pathological markers which are associated with patient survival and HCC recurrence. Liver inflammatory processes offer a large variety of systemic and pathological markers which may be exploited to improve the reliability of prognosis and decision making of liver surgeons and hepatologists. The following review aims to dissect the potential tools, targets and prognostic meaning of inflammatory markers in patients with resectable HCC. We analyze changes in circulant cellular populations and assess inflammatory biomarkers as a surrogate of impaired outcomes and provide an overview on predictive gene expression signatures including inflammatory transcriptional patterns, which are representative of poor survival in these patients.

Development and independent validation of a predictive gene expression signature based on RB1, PTEN, and TP53 in patients with metastatic hormone-sensitive metastatic prostate cancer (mHSPC).

5083 Background: Alterations on the tumor suppressor genes (TSG) RB1, PTEN and TP53 are associated with treatment resistance and aggressive clinical evolution of prostate cancer patients (pts). Here, we developed and validated a TSG gene expression signature in mHSPC pts. Methods: This is a multicenter retrospective biomarker study in mHSPC pts. Mutations on TSG were assessed by targeted sequencing, mRNA expression by nCounter platform and protein by IHC in FFPE tumor samples. Expression data of the training cohort were used to establish the cut-off applied to the other cohorts. TSGlow was considered when ≥2 out of 3 TSG presented low expression and TSGwt in the remaining cases. CHAARTED trial pts with microarray data (Hamid, Ann Oncol, 2021) were analyzed as an independent validation cohort. TSG signature was correlated with castration resistance-free survival (CRPC-FS) (primary endpoint) and overall survival (OS) by Kaplan Meier and multivariate Cox analysis. Results: 218 pts were included: 125 treated with ADT + docetaxel (ADT+D) and 93 with ADT. 160 pts from the CHAARTED dataset were analyzed (84 with ADT+D and 76 with ADT). ADT+D pts with available TSG gene expression and targeted sequencing were considered as the training cohort (n=54). Pts with low PTEN and RB1 gene expression had lower IHC expression (p<0.01). Low PTEN pts had a higher frequency of PTEN mutations (p<0.05). TSGlow (25.9%) pts had a worse CRPC-FS (14.3 vs 21.7 months (m); HR 2.2 (95% CI, 1.1 – 4.4) p=0.022). In the internal validation cohort of ADT+D-treated pts (n=71), TSGlow (9.9%) correlated with a worse CPRC-FS (11.7 vs 20 m, p=0.027) and OS (27.6 vs 58.1 m, p=0.012). TSGlow was independently associated with CPRC-FS (HR 2.7 (95% CI, 1.2 – 6.1) p=0.02) and OS (HR 3.7 (95% CI, 1.3 – 10.1) p=0.012). In the ADT cohort (TSGlow 21.5%), no differences in either CRPC-FS or OS were observed. When analyzed together the ADT+D and ADT cohorts, we observed that TSGwt pts treated with ADT+D had the largest CRPC-FS (21.6 m; p=0.014) and OS (58.1 m; p=0.029), while TSGlow pts receiving ADT+D had a CRPC-FS and OS similar to ADT-treated pts. Similarly, in the CHAARTED dataset, TSGlow (20.2%) pts treated with ADT+D had a worse CPRC-FS (14.8 vs 29.4 m, HR 2.2 (95% CI, 1.1 – 4.3) p=0.023) and OS (32.9 vs 57.4 m, HR 2.2 (95% CI, 1 – 4.5) p=0.04). TSGlow (35.5%) was not predictive of either CRPC-FS or OS in ADT-treated pts. Analyzing together ADT+D and ADT pts, TSGwt pts treated with ADT+D had the best CRPC-FS and OS (p<0.01). Conclusions: Our results suggest that a TSGlow expression may predict the lack of benefit to taxanes in mHSPC pts, which may be useful to personalize the treatment. [Table: see text]

Abstract 6574: Systematic analysis of the predictive gene expression signatures of immunotherapies across multiple cancer types using IOSig

Abstract Immune checkpoint inhibitors (ICI) have provided durable responses in a subset of patients in multiple solid tumor types. However, developing robust and reproducible predictive gene signatures for ICIs remains a challenge partly due to the lack of systematic comparisons across a large pan-cancer transcriptomics compendia. We recently developed Immuno-Oncology Signatures Explorer (IOSig), a user-friendly web tool, to allow users to query and explore predictive biomarkers from a large collection of pan-cancer patient samples treated with ICIs (> 2500 patient samples from 40 studies). Here, we investigated six solid tumors: kidney, melanoma, bladder, head and neck squamous cell, gastrointestinal, and non-small cell lung cancers tissue-based RNA-seq datasets in IOSig. The 2175 samples, from 45 cohorts, were Z-normalized with the genes from each signature being averaged to assign a sample a single score per signature. Area Under the Receiver Operating Characteristic (AUROC) curve was used to summarize the gene signatures’ ability to predict response to ICIs. We focus on the top 10 predictive gene signatures for each cancer type. There were 22 unique gene signatures from the top 10 of the six cancer types. Overall, 17 signatures were predictive in GI, NSCLC, melanoma and HNSCC. Many of these signatures were enriched in immune activation and inflammatory signatures. Conversely, the MYC, DNA damage repair, and proliferation signatures are only predictive in bladder cancer. No predictive signatures were associated with kidney cancer. Interestingly, the immune activation signatures were highly predictive in the GI cancer cohorts in our study, as the cohorts were enriched with patients having high tumor mutational burdens and microsatellite instability. In summary, we used IOSig to analyze the predictive ability of previously published gene signatures across 6 cancer types. Some of these signatures warrant further investigation in a cancer-specific manner. A table showing the AUROC value of the 22 selected gene signatures in each cancer type. Signature Melanoma Bladder Kidney NSCLC HNSC GI Cancer Average Chemokines 0.657 0.563 0.526 0.721 0.657 0.870 0.665 TIP Hot 0.657 0.571 0.544 0.737 0.592 0.866 0.661 ifng18 0.674 0.557 0.584 0.680 0.589 0.871 0.659 Ipi_Neo 0.660 0.567 0.563 0.675 0.602 0.859 0.654 Rooney 0.669 0.569 0.568 0.661 0.646 0.797 0.652 Chaurio 0.670 0.572 0.570 0.679 0.570 0.834 0.649 ifng6 0.655 0.593 0.553 0.627 0.635 0.823 0.648 ifng-effector 0.672 0.566 0.567 0.686 0.564 0.830 0.648 Roh 0.661 0.551 0.565 0.697 0.548 0.856 0.646 NRS 0.649 0.512 0.588 0.687 0.605 0.818 0.643 effector_t 0.679 0.567 0.579 0.670 0.524 0.835 0.642 impres 0.619 0.516 0.577 0.706 0.555 0.847 0.637 mMDSC 0.643 0.524 0.558 0.706 0.557 0.797 0.631 Ock 0.640 0.563 0.586 0.579 0.578 0.827 0.629 Davoli 0.665 0.546 0.569 0.655 0.514 0.822 0.629 MHC_1 0.644 0.593 0.550 0.574 0.553 0.806 0.620 MHC_2 0.661 0.533 0.554 0.659 0.437 0.802 0.608 gMDSC 0.582 0.477 0.526 0.706 0.458 0.695 0.574 DNA_damage_repair 0.432 0.598 0.555 0.383 0.499 0.361 0.471 Proliferation 0.434 0.655 0.530 0.370 0.420 0.372 0.463 Mitoscore 0.422 0.588 0.374 0.418 0.580 0.312 0.449 MYC 0.367 0.607 0.503 0.381 0.482 0.303 0.441 Citation Format: Samuel Coleman, Caroline Wheeler, Rebecca Hoyd, Louis Denko, Ching-Nung Lin, Muhammad Z. Fadlullah, Siwen Hu-Lieskovan, Christine Chung, Ahmad A. Tarhini, Daniel Spakowicz, Aik Choon Tan. Systematic analysis of the predictive gene expression signatures of immunotherapies across multiple cancer types using IOSig [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6574.

18P Systematic evaluation of published predictive gene expression signatures in pan-cancer patient cohorts treated with immune checkpoint inhibitors in a real-world setting

Previous studies developed predictive biomarkers related to immune checkpoint inihibitors (ICI) based on clinical trials cohorts. Here, we leverage the ORIEN data conducted under the Total Cancer Care protocol across 18 collaborating cancer centers to evaluate the predictive value of published gene expression signatures in real-world data (RWD) of patients with cancer treated with ICI.

Systematic evaluation of the predictive gene expression signatures of immune checkpoint inhibitors in metastatic melanoma.

Advances in immunotherapy, including immune checkpoint inhibitors (ICIs), have transformed the standard of care for many types of cancer including melanoma. ICIs have improved the overall outcome of melanoma patients; however, a significant proportion of patients suffer from primary or secondary tumor resistance. Therefore, there is an urgent need to develop predictive biomarkers to better select patients for ICI therapy. Numerous biomarkers that predict the response of melanoma to ICIs have been investigated, including biomarker signatures based on genomics or transcriptomics. Most of these predictive biomarkers have not been systematically evaluated across different cohorts to determine the reproducibility of these signatures in metastatic melanoma. We evaluated 28 previously published predictive biomarkers of ICIs based on gene expression signatures in eight previously published studies with available RNA-sequencing data in public repositories. We found that signatures related to IFN-γ-responsive genes, T and B cell markers, and chemokines in the tumor immune microenvironment are generally predictive of response to ICIs in these patients. In addition, we identified that these predictive biomarkers have higher predictive values in on-treatment samples as compared to pretreatment samples in metastatic melanoma. The most frequently overlapping genes among the top 18 predictive signatures were CXCL10, CXCL9, PRF1, RANTES, IFNG, HLA-DRA, GZMB, and CD8A. From gene set enrichment analysis and cell type deconvolution, we estimated that the tumors of responders were enriched with infiltrating cytotoxic T-cells and other immune cells and the upregulation of genes related to interferon-γ signaling. Conversely, the tumors of non-responders were enriched with stromal-related cell types such as fibroblasts and myofibroblasts, as well as enrichment with T helper 17 cell types across all cohorts. In summary, our approach of validating and integrating multi-omics data can help guide future biomarker development in the field of ICIs and serve the quest for a more personalized therapeutic approach for melanoma patients.

Abstract 3475: An 18-gene expression model predicts resistance to standard of care therapy on 3-month follow up 18FDG-PET in locally advanced cervical cancer

Abstract Introduction: As many as 30-50% of patients with locally advanced cervical cancer (LACC) experience recurrence after chemoradiation therapy (CRT), and five-year survival rates for these patients is only ~10%. While local recurrence and overall survival are useful metrics for analyzing genomic risk factors in LACC, such time to event analysis can be complicated by confounding variables which impede the elucidation of biologically relevant signals. Our group previously reported upregulation of genes from the PI3K pathway in cervical tumors with residual abnormal 18F-Fluoro-deoxy-glucose (FDG) uptake on positron emission tomography (PET) performed 3 months after CRT. Here, we analyzed whole transcriptome data using RNASeq from N=86 prospectively collected pretreatment cervix tumor biopsies to identify novel gene expression signatures associated with persistent or new FDG uptake on 3 month post-therapy FDG-PET. Objective: To identify and validate a predictive gene expression signature able to distinguish primary tumors likely to be resistant to standard of care CRT as assessed by PET scan obtained 3-months following completion of therapy. Methods: Whole tumor RNA-seq analysis was performed on 86 pre-treatment tumor specimens and compared with associated 3-month post-CRT FDG-PET. Response was dichotomized into metabolic complete response (mCR) and persistent/new disease (mP/N) defined as persistent or new areas of FDG-uptake on post-CRT PET per the nuclear medicine report. Differential expression analysis was conducted using the EdgeR R-package. Pathway enrichment analysis was conducted using reference libraries as specified and EnrichR software. Decision tree modeling, hyperparameter optimization, and k-fold cross-validation was performed using the sci-kitlearn python package. Results: We identified an 18-gene expression signature predictive for mCR vs. mP/N in patients treated with definitive CRT with an overall prospective accuracy of 0.92. This model reliably identified both classes of patients (mCR F1=0.95, precision = 0.90, recall =1.0 ; mP/N F1=0.80, precision 0.67, recall =1.0). Crowd source enrichment analysis using EnrichR identified an NRF2 driven expression program (p=0.0006), and an altered inflammatory response (0.000006) within the predictive 18 gene signature. Conclusions: We generated an 18-gene model to predict response to CRT as assessed by 3 month PET imaging. This novel expression signature further identified the NRF2 transcription factor as an important marker for resistance to CRT, consistent with previous studies showing NRF2 promotes cervical cancer growth. Our findings highlight the need for further elucidation of the NRF2 pathway in LACC, and can be used to accurately assess high risk patients who would benefit from treatment escalation. Citation Format: Michael R. Waters, Matthew Inkman, Perry W. Grigsby, Stephanie Markovina, Julie K. Schwarz, Jin Zhang. An 18-gene expression model predicts resistance to standard of care therapy on 3-month follow up 18FDG-PET in locally advanced cervical cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3475.

LungTS: A new gene expression signature for prediction of response to checkpoint inhibitors in non-small cell lung cancer.

e21143 Background: Immune checkpoint inhibitors (ICIs) are associated with durable antitumor activity in advanced non-small cell lung cancer (NSCLC), however, selection of the best candidates for this type of therapy is a challenge. PD-L1 expression, the only adequately validated predictive biomarker of response to ICIs, is imperfect, and not completely discriminatory between responders and non-responders. Gene expression profile (GEP) is a promising way to evolve in this field. The objective of this study was to identify a GEP of immune-oncology related genes, predictive of response to ICIs in NSCLC. Methods: This study analyzed two cohorts of advanced NSCLC patients treated with ICIs in any line: a discovery cohort composed of 66 Brazilian patients and a validation cohort composed of 54 Spanish patients. Clinical data were collected from medical records. Total RNA was extracted from FFPE tumor tissue. Gene expression profile (GEP) was assessed using the nCounter PanCancer IO360 panel (NanoString Technologies), which comprises 770 genes involved in the cancer immune response. LASSO regression with Cox proportional hazards approach was used to define the new gene expression signature. Results: In the discovery cohort, median age was 62.5 years (36 - 81 years). Most patients were males (57.6%), stage IV (91.9%), ECOG-PS≤1 (90.7%), smokers/former smokers (83.3%) and had adenocarcinoma histology (67.7%). The majority was treated with ICI in the first-line setting (42.4%) and nivolumab was the most frequently used drug (52.6%). At data cutoff, median follow-up for overall survival (OS - time from ICI initiation to death by any cause) was 27.7 months and for real-world progression-free survival (rwPFS - time from ICI initiation to disease progression as assessed by treating physician) was 28.2 months. The median OS and rwPFS were 17.7 months and 5.6 months, respectively. We identified a 6-gene signature (Lung Tumor Score – LungTS) that discriminate patients at low and high risk for death (median OS: 36.1 vs 15.5 months; p < 0.001) and for disease progression (median rwPFS: 10.1 vs 4.2 months; p < 0.001). Considering the expression of CD274 (PD-L1), the LungTS signature also discriminated two populations regarding OS and rwPFS (p < 0.0001). In the validation cohort, which had clinicopathological characteristics similar to the discovery cohort, the OS for high versus low LungTS was 17.0 months and 9.0 months, respectively (p = 0.004). Patients with high score also showed better rwPFS, although not statistically significant (median 10.0 versus 4.0 months; p = 0.11). Conclusions: Our study identified a novel 6-gene immune-signature predictive of response to ICI in NSCLC.

A CD8+ T Cell-Related Genes Expression Signature Predicts Prognosis and the Efficacy of Immunotherapy in Breast Cancer.

Immunotherapy has been applied to patients with breast cancer. However, only part of patients benefits from the current immunotherapy. Accurate prediction of individual response to immunotherapy can be beneficial for breast cancer management. CD8+ T cells are the main force of anti-tumor immunity. This study aimed to establish a CD8+ T cell-related gene expression signature for prediction of breast cancer prognostic and immunotherapy efficacy. RNA-seq transcriptomic data was the basics of this research. Weighted gene co-expression network analysis (WGCNA) and the least absolute shrinkage and selection operator (LASSO) Cox regression analysis established the prognostic signature. We identified 290 CD8+ T cell-related genes in the training set and established a risk-score model based on 8-genes panel (SOCS1, IL10, CAMK4, CXCL13, KIR2DS4, TESPA1, CD70 and ICAM4). Subsequently, univariate Cox regression analysis suggested that high risk-score was a risk factor for breast cancer (HR = 3.1, 95%CI 2.0-4.8, P < 0.001). In tumor microenvironment, high-risk tumors present decreased tumor infiltrating CD8+ T cells and increased M2 macrophages. The low-risk patients may benefit more from immune checkpoint blockade immunotherapy than the high-risk patients. Moreover, breast tumors which sensitive to immune checkpoint inhibitor (ICI) showed higher IL10 expression.

1836P Development of an autophagy-related gene expression signature for long term prognosis prediction in neuroblastoma patients

Neuroblastoma is one of the most malignant tumors in children. Autophagy-related genes (ARGs) play an essential role in neuroblastoma, but the expression signature has rarely been investigated in prognosis. We aimed to identify and assess prognostic ARGs signature to predict overall survival (OS) in the patients.

Abstract 654: Discovery and validation of a gene expression signature for recurrence prediction in high-risk diffuse-type gastric cancer

Abstract Background: Diffuse type gastric cancer (DGC), represented by low sensitivity to 5-fluorouracil based chemotherapy and poor prognosis, is a heterogenous malignancy in which patient subsets exhibit diverse oncological risk-profiles. The current risk stratification criteria based on the TNM staging system are inadequate in predicting the prognosis and in guiding optimal treatment in patients with DGC. Our study aimed to develop molecular biomarkers for prognostic risk-stratification in patients with DGC. Methods: We undertook a systematic and comprehensive discovery and validation effort to identify recurrence prediction biomarkers by analyzing genome-wide transcriptomic expression profiling data from 157 patients with DGC (GSE62254, GSE13861, and TCGA-STAD). Rigorous bioinformatic approaches were used to identify biomarker panels, followed by evaluation of their performance in two independent clinical cohorts (a training cohort, n=180; a validation cohort, n=74) using qRT-PCR assays. Furthermore, using Cox proportional hazard and Kaplan-Meier analyses, the clinical significance of the biomarkers was evaluated for their ability to predict recurrence individually, and when analyzed in combination with key prognostic factors. Results: Genome-wide transcriptomic profiling identified a 7-gene panel for robust prediction of recurrence in DGC patients (AUC=0.91; 95% CI: 0.83-0.96), which was successfully validated in an independent dataset (AUC=0.86; 95% CI: 0.66-0.96). Examination of 180 tissue specimens from a training cohort allowed us to establish a risk prediction model (AUC=0.78; 95% CI: 0.71-0.84), which was subsequently validated in an independent cohort of 74 GC patients (AUC=0.83; 95% CI: 0.72-0.90). The Kaplan-Meier analyses exhibited a consistently superior performance of our 7-gene risk-prediction model in the identification of high and low-risk patient subgroups. Interestingly, the performance of our gene panel to predict recurrence was significantly improved when combined with the tumor stage, in both clinical cohorts (AUC=0.83; 95% CI: 0.76-0.88 - training cohort; AUC=0.89; 95% CI: 0.80-0.95 - validation cohort). Finally, for an easier clinical translation, we established a nomogram that comprised of the 7-gene panel and the tumor stage, which robustly predicted prognosis in patients with DGC. Conclusions: We developed a novel gene expression-based risk-stratification model for predicting recurrence and the identification of high-risk patients, which could facilitate individualized treatments and improving the survival outcomes in patients suffering from DGC. Citation Format: Inseob Lee, Divya Sahu, Hoon Hur, Jeong-Hwan Yook, Byung-Sik Kim, Ajay Goel. Discovery and validation of a gene expression signature for recurrence prediction in high-risk diffuse-type gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 654.

Prognostic Cancer Gene Expression Signatures: Current Status and Challenges

Current staging systems of cancer are mainly based on the anatomical extent of disease. They need refinement by biological parameters to improve stratification of patients for tumor therapy or surveillance strategies. Thanks to developments in genomic, transcriptomic, and big-data technologies, we are now able to explore molecular characteristics of tumors in detail and determine their clinical relevance. This has led to numerous prognostic and predictive gene expression signatures that have the potential to establish a classification of tumor subgroups by biological determinants. However, only a few gene signatures have reached the stage of clinical implementation so far. In this review article, we summarize the current status, and present and future challenges of prognostic gene signatures in three relevant cancer entities: breast cancer, colorectal cancer, and hepatocellular carcinoma.

DNA Repair-Based Gene Expression Signature and Distinct Molecular Subtypes for Prediction of Clinical Outcomes in Lung Adenocarcinoma.

Objective: To conduct a robust prognostic gene expression signature and characterize molecular subtypes with distinct clinical characteristics for lung adenocarcinoma (LUAD).Methods: Based on DNA repair genes from the GSEA database, a prognostic signature was conducted in the TCGA-LUAD training set via univariate and multivariate cox regression analysis. Its prediction power was validated by overall survival analysis, relative operating characteristic (ROC) curves and stratification analysis in the GSE72094 verification set. Involved pathways in the high- and low-risk groups were analyzed by GSEA. A nomogram was built based on the signature and clinical features and its performance was assessed by calibration plots. LUAD samples were clustered via the ConsensusClusterPlus package. The differences in clinical outcomes, single nucleotide polymorphism (SNP) and sensitivity to chemotherapy drugs between molecular subtypes were analyzed.Results: A 13-DNA repair gene-signature was constructed for LUAD prognosis. Following validation, it can robustly and independently predict patients' clinical outcomes. The GSEA results exhibited the differences in pathways between high- and low- risk groups. A nomogram combining the signature and stage could accurately predict 1-, 3-, and 5-year survival probability. Two distinct molecular subtypes were characterized based on DNA repair genes. Patients in the Cluster 2 exhibited a worse prognosis and were more sensitive to common chemotherapy than those in the Cluster 1.Conclusion:This study proposed a 13-DNA repair gene-signature as a prognostic factor for LUAD patients, which can independently predict clinical outcomes by complement of the stage. Moreover, we characterized two LUAD subtypes with distinct clinical outcomes, somatic gene mutations, and drug sensitivity in cancer based on DNA repair genes.

Distinct clinical and biological characteristics of acute myeloid leukemia with higher expression of long noncoding RNA KIAA0125

Expression of long non-coding RNA KIAA0125 has been incorporated in various gene expression signatures for prognostic prediction in acute myeloid leukemia (AML) patients, yet its functions and clinical significance remain unclear. This study aimed to investigate the clinical and biological characteristics of AML bearing different levels of KIAA0125. We profiled KIAA0125 expression levels in bone marrow cells from 347 de novo AML patients and found higher KIAA0125 expression was closely associated with RUNX1 mutation, but inversely correlated with t(8;21) and t(15;17) karyotypes. Among the 227 patients who received standard chemotherapy, those with higher KIAA0125 expression had a lower complete remission rate, shorter overall survival (OS) and disease-free survival (DFS) than those with lower expression. The prognostic significance was validated in both TCGA and GSE12417 cohorts. Subgroup analyses showed that higher KIAA0125 expression also predicted shorter DFS and OS in patients with normal karyotype or non-M3 AML. In multivariable analysis, higher KIAA0125 expression remained an adverse risk factor independent of age, WBC counts, karyotypes, and mutation patterns. Bioinformatics analyses revealed that higher KIAA0125 expression was associated with hematopoietic and leukemic stem cell signatures and ATP-binding cassette transporters, two predisposing factors for chemoresistance.

Gene Networks Constructed Through Simulated Treatment Learning can Predict Proteasome Inhibitor Benefit in Multiple Myeloma.

Proteasome inhibitors are widely used in treating multiple myeloma, but can cause serious side effects and response varies among patients. It is, therefore, important to gain more insight into which patients will benefit from proteasome inhibitors. We introduce simulated treatment learned signatures (STLsig), a machine learning method to identify predictive gene expression signatures. STLsig uses genetically similar patients who have received an alternative treatment to model which patients will benefit more from proteasome inhibitors than from an alternative treatment. STLsig constructs gene networks by linking genes that are synergistic in their ability to predict benefit. In a dataset of 910 patients with multiple myeloma, STLsig identified two gene networks that together can predict benefit to the proteasome inhibitor, bortezomib. In class "benefit," we found an HR of 0.47 (P = 0.04) in favor of bortezomib, while in class "no benefit," the HR was 0.91 (P = 0.68). Importantly, we observed a similar performance (HR class benefit, 0.46; P = 0.04) in an independent patient cohort. Moreover, this signature also predicts benefit for the proteasome inhibitor, carfilzomib, indicating it is not specific to bortezomib. No equivalent signature can be found when the genes in the signature are excluded from the analysis, indicating that they are essential. Multiple genes in the signature are linked to working mechanisms of proteasome inhibitors or multiple myeloma disease progression. STLsig can identify gene signatures that could aid in treatment decisions for patients with multiple myeloma and provide insight into the biological mechanism behind treatment benefit.

Ex Vivo Drug Sensitivity and Functional Genomics Platform Identifies Novel Combinations Targeting Intrinsic and Extrinsic Apoptotic Signaling Pathways in Multiple Myeloma

Introduction: Despite some long-term remissions, eventual drug resistance in most patients remains a critical obstacle in the treatment of multiple myeloma (MM). The development of new drugs/drug combinations with novel mechanisms of action are needed for continued improvement in patient outcomes. Initiation of tumor cell death via activation of the intrinsic (mitochondrial) and/or extrinsic (death receptor) apoptotic signaling pathways has been shown to be an effective therapeutic strategy in MM. Venetoclax (Ven) is a selective, small-molecule inhibitor of BCL-2 that exhibits clinical activity in MM cells, particularly in patients harboring the t(11;14) translocation. Navitoclax (Nav) is a small-molecule that targets multiple antiapoptotic BCL-2 family proteins, including BCL-XL, BCL-2, and BCL-W to initiate the intrinsic apoptotic pathway. Eftozanermin alfa (Eftoza) is a novel, second generation TRAIL receptor agonist that induces cell death via death receptor pathways and is under investigation in multiple solid and heme malignancies. In addition, the pan-BET inhibitor mivebresib (Miv) and the BDII selective BET inhibitor ABBV-744 have shown synergistic activity with Ven in cell line models of multiple heme malignancies. Results reported here describe ex vivo drug sensitivities and functional genomic analyses of Ven, Nav, Eftoza, Miv, and ABBV-744 alone or in combination with standard-of-care agents, including bortezomib, carfilzomib, panobinostat, daratumumab, or pomalidomide. Methods: A high-throughput ex vivo drug screening assay using a coculture system of bone marrow (BM)-derived MM and stromal cells was used to assess the sensitivity of MM patient tumor cells (Figure 1A). Paired whole exome sequencing (WES) and RNA sequencing (RNA-seq) analyses were performed. Results: Primary MM patient specimens (n=52) were evaluated in the ex vivo platform, including treatment-naïve, early relapse (1-3 prior lines), and late relapse (4-8 prior lines) patients treated with proteasome inhibitors, immunomodulatory drugs, and monoclonal antibodies. As expected, t(11;14)-positive MM patient specimens were more sensitive than wildtype to Ven ex vivo (D AUC, -18.6, P=0.002), however MM cells harboring amp(1q) were more resistant than wildtype (D AUC, +5.07, P=0.032), suggesting MCL1 (1q21 gene locus) is a key resistance factor to Ven single-agent activity in MM. Gene set enrichment analysis identified B-cell receptor signaling (normalized enrichment score (NES), 1.96, adjusted P=0.010) and MYC pathway (NES, 1.95, adjusted P=0.010) overexpression as predictors of increased sensitivity to Ven ex vivo. A t(11;14) gene expression signature was also generated using a penalized regression model approach in an additional MMWG/ORIEN MM patient cohort (n=155). The t(11;14) predictive gene expression signature was confirmed by correlation with Ven AUC in the ex vivo model. Additional pathway analyses were performed to identify potential predictive markers of sensitivity/resistance for each single agent and drug combination. Although ex vivo activity of Nav was higher in t(11;14) specimens compared to non-t(11;14) (D AUC, -17.8, P=0.011), ex vivo activity in non-t(11;14) specimens was also observed, indicating additional anti-MM activity by cotargeting of BCL-XL and BCL-2. Both Miv and ABBV-744 showed single-agent activity ex vivo, however Miv demonstrated higher activity (median LD50=88.4nM), suggesting that pan-BET inhibition is more effective than BDII-specific BET inhibition in MM. Finally, a novel drug-combination effect analysis was used that identified novel synergistic ex vivo combinations including Ven and panobinostat (P=0.0013) and Eftoza with bortezomib (P=1.8E-7) or carfilzomib (P=7E-4). Additionally, single-agent induction of macrophage-mediated phagocytosis was observed in both Ven and daratumumab, which was synergistic when the 2 drugs were combined (Figure 1B). Conclusion: An ex vivo functional genomic screen of MM patient specimens demonstrated the usefulness of this approach to identify candidate drugs and potential predictive biomarkers for continued evaluation in clinical trials. This approach confirmed known mechanisms of drug sensitivity and identified new ones, including a novel characterized immune-mediated synergy between Ven and daratumumab, and potential combination strategy for Eftoza and proteasome inhibitors. Figure 1 Disclosures Siqueira Silva: Karyopharm: Research Funding; NIH/NCI: Research Funding; AbbVie: Research Funding. Kulkarni:M2GEN: Current Employment. Mitchell:AbbVie: Other: payment for bioinformatics analysis, Research Funding; M2GEN: Current Employment, Research Funding. Dai:Cygnal Therapeutics: Current Employment; M2GEN: Ended employment in the past 24 months. Hampton:M2GEN: Current Employment. Lu:AbbVie: Current Employment, Current equity holder in publicly-traded company. Modi:AbbVie: Current Employment, Other: may own stock or stock options. Motwani:AbbVie: Current Employment, Current equity holder in publicly-traded company. Harb:AbbVie: Current Employment, Other: may hold stock or stock options. Ross:AbbVie: Current Employment, Current equity holder in publicly-traded company. Shain:Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; GlaxoSmithKline: Speakers Bureau; Sanofi/Genzyme: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Karyopharm: Research Funding, Speakers Bureau; AbbVie: Research Funding; Takeda: Honoraria, Speakers Bureau; Janssen: Honoraria, Speakers Bureau; Amgen: Speakers Bureau; Adaptive: Consultancy, Honoraria; BMS: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau. OffLabel Disclosure: While this is a preclinical study, venetoclax for treatment of multiple myeloma is not an approved indication

Abstract 832: Construction and optimization of gene expression signatures for prediction of survival in two-arm clinical trials

Abstract Gene expression signatures for the prediction of differential survival of patients undergoing anti-cancer therapies are of great interest because they can be used to prospectively stratify patients entering new clinical trials, or to determine optimal treatment for patients in more routine clinical settings. Unlike prognostic signatures however, predictive signatures require training set data from clinical studies with at least two treatment arms, and the methodology for constructing and optimizing predictive signatures has been less prominently explored. Focusing on the “use case” of a two-arm clinical trial for metastatic colorectal cancer (CRC) patients treated with the anti-angiogenic molecule aflibercept, we present tools for the construction of signatures predictive of progression-free survival based on cross-validated multivariate Cox models. The work is organized around a specific approach called subtype correlation, which leverages a priori knowledge of molecular subtypes of tumors for a given indication. The tools and concepts presented here include the so-called differential log-hazard ratio, the survival scatter plot, the hazard ratio receiver operating characteristic, the area between curves and the patient selection matrix. In the CRC use case, the resulting signature stratifies the patient population into “sensitive” and “relatively-resistant” groups achieving a more than two-fold difference in the aflibercept-to-control hazard ratios across signature-defined patient groups. Through cross-validation and resampling the probability of generalization of the signature to similar CRC data sets is predicted to be high. The tools presented here should be of general use for building and using predictive multivariate signatures in oncology and in other therapeutic areas. Citation Format: Joachim Theilhaber, Marielle Chiron, Jennifer Dreymann, Donald Bergstrom, Jack Pollard. Construction and optimization of gene expression signatures for prediction of survival in two-arm clinical trials [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 832.

Construction and optimization of gene expression signatures for prediction of survival in two-arm clinical trials

BackgroundGene expression signatures for the prediction of differential survival of patients undergoing anti-cancer therapies are of great interest because they can be used to prospectively stratify patients entering new clinical trials, or to determine optimal treatment for patients in more routine clinical settings. Unlike prognostic signatures however, predictive signatures require training set data from clinical studies with at least two treatment arms. As two-arm studies with gene expression profiling have been rarer than similar one-arm studies, the methodology for constructing and optimizing predictive signatures has been less prominently explored than for prognostic signatures.ResultsFocusing on two “use cases” of two-arm clinical trials, one for metastatic colorectal cancer (CRC) patients treated with the anti-angiogenic molecule aflibercept, and the other for triple negative breast cancer (TNBC) patients treated with the small molecule iniparib, we present derivation steps and quantitative and graphical tools for the construction and optimization of signatures for the prediction of progression-free survival based on cross-validated multivariate Cox models. This general methodology is organized around two more specific approaches which we have called subtype correlation (subC) and mechanism-of-action (MOA) modeling, each of which leverage a priori knowledge of molecular subtypes of tumors or drug MOA for a given indication. The tools and concepts presented here include the so-called differential log-hazard ratio, the survival scatter plot, the hazard ratio receiver operating characteristic, the area between curves and the patient selection matrix. In the CRC use case for instance, the resulting signature stratifies the patient population into “sensitive” and “relatively-resistant” groups achieving a more than two-fold difference in the aflibercept-to-control hazard ratios across signature-defined patient groups. Through cross-validation and resampling the probability of generalization of the signature to similar CRC data sets is predicted to be high.ConclusionsThe tools presented here should be of general use for building and using predictive multivariate signatures in oncology and in other therapeutic areas.