Gene Expression Classifier Research Articles

Combined molecular and histologic end points inform cancer risk estimates for thyroid nodules classified as atypia of undetermined significance.

Thyroid nodules classified as atypia of uncertain significance (AUS) on fine-needle aspiration cytology are heterogeneous. Prior studies reported a higher risk of noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP)/cancer among AUS nodules that had cytologic (AUS-C) versus architectural (AUS-A) atypia; however, such studies were generally confined to resected cohorts, introducing bias into risk calculations. The authors hypothesized that combined histologic and molecular end points would permit clinically meaningful calculations of NIFTP/malignancy risk among AUS nodules. The study consisted of 279 thyroid nodules classified as AUS on initial fine-needle aspiration and tested by the Afirma Gene Expression Classifier (GEC) between June 2013 and October 2017. Results of GEC testing and histopathologic diagnoses were stratified by AUS classifiers. The AUS-A category was further subclassified as 1) hypocellular microfollicular or 2) cellular with mixed but predominantly microfollicular architecture. NIFTP/cancer risk was calculated for each subgroup, with the inclusion of unresected nodules that had benign GEC results as low-risk end points comparable to histologically benign nodules. When only histologic end points were considered, there was no difference in NIFTP/cancer risk (25% vs 23%; P = .82). By using molecular and histologic end points, AUS cases with cytologic atypia trended toward higher NIFTP/cancer risk than AUS-A cases (14% vs 6%; P = .06). Furthermore, AUS-A cases showed a trend toward lower NIFTP/cancer risk for hypocellular microfollicular aspirates (3%) compared with cellular samples that had mixed/predominantly microfollicular architecture (13%; P = .18). The inclusion of unresected benign GEC nodules in risk-of-malignancy calculations provides more accurate results, which may be helpful for informing patient management as well as quality improvement in the cytopathology laboratory.

Abstract LB018: Cloud-based biological domain knowledge derived gene expression classifier to predict the cancer clinical outcome

Abstract We report the advances of a cloud-based software utilizing advanced algorithms to identify the molecular cancer drivers and its signal transduction. To test, a classifier based on the NFκB signaling pathway gene expression profiles that predicts the clinical therapeutic response to chemo-radiation therapy of Squamous Cell Carcinomas of the Head and Neck (SCCHN) patients are described here. Expression of HPV16, p16 and 84 NFκB signaling pathway genes were determined using real time RT PCR in SCCHN patient samples and SCCHN cell lines. Gene expression was correlated to chemo-radiation response in cell lines and the model was blindly tested in SCCHN patients to predict their clinical outcome to chemo-radiation. This tool intends to fuse biological domain knowledge by way of integrating disease targets, cell line experiments towards creating a gene expression classifier for clinical outcome. The gene expression-based predicted clinical outcome compared with patients actual observed clinical outcome correctly identified the clinical response in 73% of the cases. Such work will allow design of tailored individualized clinical treatment strategies to target NFκB inhibition using biological modifying agents with or without chemo/radiation therapy. This will help to develop several targets to directly inhibit the NFκB activity or inhibit the downstream pivotal targets. The significance of the classifier would be the ability to use the biologic attributes of a tumor to tailor specific treatment protocols for each individual patient. This would not only spare the chemo-radiation resistant patients the ill effects of the therapy but will also decrease the delay in effective treatment among the resistant patients. The long-term objective is to use such pro-survival driver genes in other cancers; SMAD, TGFβ in pancreatic cancer, and AR in prostate cancer. This tool (MeghaOncomine) intends to fuse biological domain knowledge by way of integrating cancer targets, cell line experiments towards creating a gene expression classifier for clinical outcome. Citation Format: Rao V. Papineni, Seema Gupta, Aravindan Natarajan, Ram Papineni, Chandra Kasinathan. Cloud-based biological domain knowledge derived gene expression classifier to predict the cancer clinical outcome [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 LB018.

The Cancer SENESCopedia: A delineation of cancer cell senescence.

SummaryCellular senescence is characterized as a stable proliferation arrest that can be triggered by multiple stresses. Most knowledge about senescent cells is obtained from studies in primary cells. However, senescence features may be different in cancer cells, since the pathways that are involved in senescence induction are often deregulated in cancer. We report here a comprehensive analysis of the transcriptome and senolytic responses in a panel of 13 cancer cell lines rendered senescent by two distinct compounds. We show that in cancer cells, the response to senolytic agents and the composition of the senescence-associated secretory phenotype are more influenced by the cell of origin than by the senescence trigger. Using machine learning, we establish the SENCAN gene expression classifier for the detection of senescence in cancer cell samples. The expression profiles and senescence classifier are available as an interactive online Cancer SENESCopedia.

Discovery of an Immune-Driven Gene Expression Signature That Predicts for Clinical Benefit to CC-122 in Relapsed or Refractory Diffuse Large B-Cell Lymphoma (R/R DLBCL)

Introduction: CC-122, a small-molecule cereblon-binding agent, degrades the hematopoietic transcription factors Ikaros (IKZF1) and Aiolos (IKZF3) by modulating the CRL4CRBN E3 ligase complex and shows clinical activity in patients (pts) with R/R DLBCL. In preclinical models, CC-122 has activity in ABC and GCB DLBCL cells, which differentiates it from lenalidomide (primarily active in the ABC subtype). Thus, cell of origin is not an appropriate selection strategy for CC-122. Herein described are 2 DLBCL patient phenotypes, identified de novo from public gene expression data, which describe immune infiltration and increased tumor cellularity. When incorporated into a gene expression classifier that distinguishes between them, and applied to gene expression data from samples collected in a trial of CC-122 monotherapy in R/R DLBCL (NCT01421524; CC-122-ST-001), these phenotypes enrich for improved response and prolonged PFS.Methods: Publicly available raw Affymetrix expression profiles for 414 newly diagnosed DLBCL not otherwise specified (NOS) pts (GEO accession: GSE10846; Lenz et al, NEJM 2008) were normalized and batch-corrected, prior to transformation of the resulting expression matrix using non-negative matrix factorization (NMF). Unsupervised analysis of the output revealed 2 major phenotypes, which were interpreted via a combination of methods and prior biological knowledge, including GSEA with GO Biological Processes and published gene signatures. The patient gene expression profiles were separated into 2 subgroups based on the phenotypes observed. Supervised classification was applied to demonstrate their predictability, and to identify a subset of discriminant genes that were used to design a custom gene expression panel on the NanoString nCounter platform. An adapted Nearest Template Prediction algorithm (Hoshida Y, PLoS 2010) was applied to the same Affymetrix data and candidate gene-set to separate the 2 sub-groups and enable the classification of NanoString profiles to select a patient phenotype potentially enriched for positive outcome to CC-122.Results: Based on the Lenz 2008 dataset, biological interpretation of the first phenotype identified pathways involved in the immune system and inflammatory response (phenotype 1), and the second phenotype was enriched for DNA replication/repair and cell cycle and RNA processing, indicating a proliferative signal and potentially higher B-cell content (phenotype 2). Phenotype 1 pts were also enriched in the 2 stromal signatures reported by Lenz et al, reflecting tumor infiltration by cells of the monocytic lineage and endothelial cells. Application of multiple immune deconvolution methods to the Lenz 2008 dataset identified an enrichment of T cells and macrophages in phenotype 1 and high B-cell content in phenotype 2. These observations yielded the hypothesis that differential contribution of these molecular phenotypes may predict pts who receive greatest benefit from CC-122. The gene expression classifier developed to separate pts in the Lenz dataset according to dominant phenotype was applied to gene expression data from Celgene trial CC-122-ST-001 to predict pts as phenotypic subgroup 1 (classifier-positive pts) or phenotypic subgroup 2 (classifier-negative pts). Initial observations in H&E staining confirmed significant differences (P <.01) in tumor cellularity (B-cell content) for classifier-positive (75%) vs classifier-negative (92%) pts (median 90 [SD 29.3; MAD 13.3] vs 98 [SD 13.7; MAD 1.5]), which reinforces the original interpretation of phenotypes discovered in the Lenz dataset. Application of the gene expression classifier to pre-treatment nCounter profiles from third-line or greater R/R DLBCL NOS pts in Celgene trial CC-122-ST-001 revealed an ORR of 34.8% and mPFS of 186 days for classifier-positive pts vs an ORR of 21.7% and mPFS of 49 days for classifier-negative pts (mPFS HR = 0.43; P=.02). No significant differences in overall survival were observed in first-line DLBCL pts treated with immunochemotherapy R-CHOP, suggesting that the classifier is not prognostic.Conclusions: A novel gene expression classifier has been developed that predicts response and prolonged PFS to CC-122 in R/R DLBCL NOS but has no prognostic value. This classifier will be explored in future trials as an enrichment strategy for pts most likely to benefit clinically from CC-122. DisclosuresRisueño:Celgene: Employment, Patents & Royalties: US15273205. Hagner:Celgene: Employment, Equity Ownership. Fontanillo:Celgene: Employment, Equity Ownership. Djebbari:Celgene: Employment. Parker:Celgene: Consultancy. Drew:Celgene Corp.: Consultancy. Towfic:Celgene Corporation: Employment, Equity Ownership; Immuneering Corporation: Equity Ownership. Pourdehnad:Celgene Corporation: Employment, Equity Ownership. Gandhi:Celgene Corporation: Employment, Equity Ownership. Trotter:Celgene Corporation: Equity Ownership; Celgene Institute for Translational Research Europe: Employment.

CC-122 DLBCL-001: Phase Ib Study of CC-122 Plus Rituximab in Patients with Chemo-Refractory Diffuse Large B-Cell Lymphoma (DLBCL)

CC-122 DLBCL-001: Phase Ib Study of CC-122 Plus Rituximab in Patients with Chemo-Refractory Diffuse Large B-Cell Lymphoma (DLBCL)

A Novel Gene Expression Classifier Enriches for Single Agent Clinical Activity of CC-122, a Cereblon Modulator, Administered Orally to Relapsed or Refractory Diffuse Large B-Cell Lymphoma Subjects: Results from the Phase I CC-122-ST-001 Study

A Novel Gene Expression Classifier Enriches for Single Agent Clinical Activity of CC-122, a Cereblon Modulator, Administered Orally to Relapsed or Refractory Diffuse Large B-Cell Lymphoma Subjects: Results from the Phase I CC-122-ST-001 Study

Performance of Afirma genomic sequencing classifier vs gene expression classifier in Bethesda category III thyroid nodules: An institutional experience

Afirma gene expression classifier (GEC) is an adjunct to thyroid fine needle aspiration shown to improve pre-operative risk assessment and reduce unnecessary surgery of indeterminate thyroid nodules. Genomic sequencing classifier (GSC) is a newer version aiming to improve specificity and positive predictive value (PPV) of Afirma testing. There are limited studies comparing GSC vs GEC. This study was undertaken to compare these classifiers in terms of diagnostic performance and effect on clinical management of indeterminate thyroid nodules. The study cohort consisted of patients with thyroid nodules that had a recurrent cytologic diagnosis of atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS) and were tested by either GEC or GSC. Patient demographics, nodule size, and clinical follow-up were recorded. Benign call rate (BCR) of Afirma testing, rate of subsequent surgery (RSS), rate of histology-confirmed malignancy (RHM), as well as diagnostic sensitivity, specificity, PPV, negative predicative value (NPV), and accuracy were calculated and compared between GSC and GEC cohorts. Among 264 AUS/FLUS thyroid nodules, 127 and 137 were tested with GEC and GSC, respectively. Compared to GEC, GSC demonstrated increased BCR (77.3% vs 52%), decreased RSS (31.4% vs 51.2%), greater RHM (29% vs 9.8%) associated with a suspicious Afirma result, as well as improved specificity (82.8% vs 54.5%), PPV (29% vs 9.8%), and diagnostic accuracy (83.9% vs 56.7%), while maintaining high sensitivity and NPV. Afirma GSC substantially improved BCR, RSS, RHM, and diagnostic performance, enhancing appropriate triage and thereby helped avoid unnecessary surgery in AUS/FLUS thyroid nodules.

Targeted gene expression classifier identifies pediatric T-cell acute lymphoblastic leukemia (T-ALL) patients at high risk for end induction minimal residual disease positivity.

10002 Background: The heterogeneity of T-ALL has hindered biomarker identification and limited biology-based risk stratification. Historically, minimal residual disease (MRD) has been the strongest predictor of poor outcomes. However, stratification by MRD does not allow for risk-adapted therapy early in treatment, which may induce deeper remissions and decrease risk of relapse. We hypothesized that gene expression profiling at diagnosis may have prognostic value in identifying high risk patients. Methods: We analyzed RNA-seq data from 189 diagnostic samples from the Children’s Oncology Group (COG) AALL0434 trial. Using leave-one-out cross-validation, we identified a set of genes that optimally differentiated MRD+ and MRD- samples. We then derived a risk score (RS) that indicates a probability of being MRD+ for a given gene expression pattern. Finally, we validated this model in an independent cohort of COG AALL1231 samples. Results: The AALL0434 early T-cell precursor (ETP) samples (n = 19), which have high rates of MRD+, had the highest RS, with an average of 81.3 (SD 18.7), versus 24.9 (SD 22.7) for non-ETPs (n = 146). Intriguingly, non-ETPs with RS &gt; 50 had a gene expression pattern that mirrored ETPs and was distinct from the remaining non-ETPs. In this RS &gt; 50 non-ETP cohort, 80% were MRD+, versus 20% of the &lt; 50 cohort (p &lt; 0.0001). When applied to 31 diagnostic non-ETP samples from COG AALL1231, 57% of the RS &gt; 50 cohort were MRD+, versus 17% of the RS &lt; 50 cohort (p = 0.05). Importantly, AALL0434 used prednisone during induction, while AALL1231 used dexamethasone, indicating that the predictive value is independent of the induction steroid. Finally, we converted our model to the customizable Nanostring nCounter platform by analyzing 96 AALL0434 samples on the Nanostring assay. The Nanostring data closely recapitulated the RNA-seq data, with a tight correlation between the resulting RS (concordance correlation coefficient = 0.91). Conclusions: We have developed a gene expression classifier that differentiates a subset of non-ETP T-ALLs with an ETP-like gene expression pattern and a high risk of MRD+, and have adapted the classifier to a clinically tractable targeted platform. Identification of this high-risk subset at diagnosis has the potential to facilitate risk-adapted trials to evaluate the utility of novel or more intensive therapies aimed at improving clinical outcomes.

Analysis of the MOSAIC correlative cancer database integrating molecular cancer classification and tumor profiling to identify targeted treatment options for metastatic cancer.

3121 Background: Unless tumor type and genetic alterations can be identified, metastatic cancer patients with unknown or uncertain diagnoses may be limited to empiric chemotherapy. The 92-gene assay (CancerTYPE ID) is a validated gene expression classifier of 50 tumor types and subtypes for patients with cancer of unknown primary (CUP) or ambiguous diagnoses. Multimodal molecular biomarker testing by next-generation sequencing (NGS), tumor mutational burden (TMB), fluorescent in situ hybridization (FISH), microsatellite instability (MSI), and immunohistochemistry (IHC) can identify genetic targets. A database integrating tumor typing with biomarker analysis in metastatic cases was utilized to identify the most prevalent genetic alterations by tumor type. Methods: MOSAIC (Molecular Synergy to Advance Individualized Cancer Care) is an IRB-approved database of cases with CancerTYPE ID testing plus multimodal biomarker testing. The current study determined molecular tumor type followed by molecular profiling by NGS for up to 323 genes, (NeoTYPE profiles, Neogenomics). Results: Tumor type was determined in 1992 of 2151 cases (92.7%), comprised of 27 different tumor types. 72% of cases were comprised of the 7 tumor types shown in the table,which also shows the frequency of the 10 most commonly mutated genes by tumor type. Bolded are genes with actionable genetic mutations for which FDA-approved therapies are not currently indicated in the identified tumor type. Conclusions: Precise targeted treatment for many patients with CUP or ambiguous diagnoses requires accurate diagnosis of the cancer origin combined with multimodal molecular testing to identify actionable genetic alterations in the appropriate cellular context. Future studies will evaluate additional biomarker profiles, including TMB, FISH, MSI, and IHC for cases in the MOSAIC database.[Table: see text]

Integration of molecular cancer classification and next-generation sequencing to identify metastatic patients eligible for PARP inhibitors.

e15080 Background: Olaparib, rucaparib, and niraparib are 3 poly-ADP-ribose polymerase inhibitors (PARPi) approved by the FDA for ovarian, breast, pancreatic, prostate, fallopian tube and peritoneal cancers with BRCA mutations. Several ongoing clinical trials aim to determine the efficacy of PARPi in various other cancer types, including specific cancer subtypes, such as clear cell renal cell carcinoma and cholangiocarcinoma either as monotherapy or combination therapy; however, eligibility for PARPi therapy requires the identification of the primary tumor type and confirmation of BRCA mutation. The 92-gene assay (CancerTYPE ID) is a validated gene expression classifier of 50 tumor types and subtypes for metastatic patients with unknown or uncertain diagnoses. Multimodal biomarker testing, including next-generation sequencing (NGS), enables identification of actionable biomarkers to guide targeted therapy selection. In the current study, a database of metastatic cases that integrates tumor type with biomarker analysis was characterized to identify those eligible for PARPi treatment. Methods: MOSAIC (Molecular Synergy to Advance Individualized Cancer Care) is an IRB-approved, de-identified database of cases submitted for CancerTYPE ID testing with tissue-guided multimodal biomarker testing by NGS, including tumor mutational burden (TMB) fluorescent in situ hybridization (FISH), and microsatellite instability (MSI), and immunohistochemistry (IHC) (NeoTYPE profiles, Neogenomics). For the current study, metastatic cancers classified as ovarian, breast, pancreatic, or prostate were identified in the database, followed by NGS analysis to detect mutations in BRCA1 or BRCA2. Results: The current analysis included 2151 CancerTYPE ID cases, from which 71 ovarian, 47 breast, 12 pancreatic and 15 prostate cancer cases were identified. Out of 46 cases of ovarian cancer with molecular biomarker results, NGS identified 7 (15.3%) cases with BRCA1 mutation and 4 (8.7%) cases with BRCA2 mutation. Additionally, 4 (10.5%) cases with BRCA1 mutation and 1 (2.6%) case with BRCA2 mutation out of 38 cases of breast cancer with BRCA results were detected. No cases of prostate cancer or pancreatic cancer with mutations in BRCA1 or BRCA2 were detected. Conclusions: These findings in metastatic patients demonstrate the clinical utility of tumor type identification combined with molecular biomarker profiling, leading to additional options for patients with advanced disease. Specifically, analysis of the MOSAIC database identified a subset of patients with metastatic cancers eligible for PARPi therapy based on tumor type and BRCA mutation status. As new and approved PARPi are evaluated for efficacy in additional tumor types, patients can be identified that may be eligible for these targeted cancer drugs.

Repeat Fine Needle Aspiration Cytology Refines the Selection of Thyroid Nodules for Afirma Gene Expression Classifier Testing.

Background: Molecular testing (MT) refines risk stratification for thyroid nodules that are indeterminate for cancer by fine needle aspiration (FNA) cytology. Criteria for selecting nodules for MT vary and remain largely untested, raising questions about the best strategy for maximizing the usefulness of MT while minimizing the harms of overtesting. We used a unique data set to examine the effects of repeat FNA cytology-based criteria for MT on management decisions and nodule outcomes. Methods: This was a study of adults (age 25-90 years; 281 women and 72 men) with cytologically indeterminate (Bethesda III/IV) thyroid nodules who underwent repeat FNA biopsy and Afirma Gene Expression Classifier (GEC) testing (N = 363 nodules from 353 patients) between June 2013 and October 2017 at a single institution, with follow-up data collected until December 2019. Subgroup analysis was performed based on classification of repeat FNA cytology. Outcomes of GEC testing, clinical/sonographic surveillance of unresected nodules, and histopathologic diagnoses of thyroidectomies were compared between three testing approaches: (i) Reflex (MT sent on the basis of the initial Bethesda III/IV FNA), (ii) SemiRestrictive (MT sent if repeat FNA is Bethesda I-IV), and (iii) Restrictive (MT sent only if repeat FNA is Bethesda III/IV) testing approaches. Results: Restricting MT to nodules that remain Bethesda III/IV on repeat FNA would have missed 4 low-risk cancers and 3 noninvasive follicular thyroid neoplasms with papillary-like nuclear features (NIFTP) (collectively 2% of the test population) but would have avoided diagnostic surgery for 42 benign nodules (12% of the test population). The Restrictive testing strategy was more specific (delta 0.126 confidence interval [CI 0.093 to 0.159] and 0.129 [CI 0.097 to 0.161], respectively) but less sensitive (delta -0.339 [CI -0.424 to -0.253] and -0.340 [CI -0.425 to -0.255], respectively) than the Reflex and SemiRestrictive approaches for detecting NIFTP or cancer. Conclusions: Repeat FNA cytology can guide the selection of cytologically indeterminate thyroid nodules that warrant MT. The Restrictive model of performing Afirma GEC only on nodules with two separate biopsies showing Bethesda III/IV cytology would reduce the rate of diagnostic surgery for histologically benign nodules while missing only rare low-risk tumors. Given the low but nontrivial risks of thyroidectomy, the higher specificity of the Restrictive testing approach disproportionately outweighs the potential harms.

Pan-cancer Transcriptomic Predictors of Perineural Invasion Improve Occult Histopathologic Detection.

Perineural invasion (PNI) is associated with aggressive tumor behavior, recurrence, and metastasis, and can influence the administration of adjuvant treatment. However, standard histopathologic examination has limited sensitivity in detecting PNI and does not provide insights into its mechanistic underpinnings. A multivariate Cox regression was performed to validate associations between PNI and survival in 2,029 patients across 12 cancer types. Differential expression and gene set enrichment analysis were used to learn PNI-associated programs. Machine learning models were applied to build a PNI gene expression classifier. A blinded re-review of hematoxylin and eosin (H&E) slides by a board-certified pathologist helped determine whether the classifier could improve occult histopathologic detection of PNI. PNI associated with both poor overall survival [HR, 1.73; 95% confidence interval (CI), 1.27-2.36; P < 0.001] and disease-free survival (HR, 1.79; 95% CI, 1.38-2.32; P < 0.001). Neural-like, prosurvival, and invasive programs were enriched in PNI-positive tumors (P adj < 0.001). Although PNI-associated features likely reflect in part the increased presence of nerves, many differentially expressed genes mapped specifically to malignant cells from single-cell atlases. A PNI gene expression classifier was derived using random forest and evaluated as a tool for occult histopathologic detection. On a blinded H&E re-review of sections initially described as PNI negative, more specimens were reannotated as PNI positive in the high classifier score cohort compared with the low-scoring cohort (P = 0.03, Fisher exact test). This study provides salient biological insights regarding PNI and demonstrates a role for gene expression classifiers to augment detection of histopathologic features.

Airway gene-expression classifiers for respiratory syncytial virus (RSV) disease severity in infants

BackgroundA substantial number of infants infected with RSV develop severe symptoms requiring hospitalization. We currently lack accurate biomarkers that are associated with severe illness.MethodWe defined airway gene expression profiles based on RNA sequencing from nasal brush samples from 106 full-tem previously healthy RSV infected subjects during acute infection (day 1–10 of illness) and convalescence stage (day 28 of illness). All subjects were assigned a clinical illness severity score (GRSS). Using AIC-based model selection, we built a sparse linear correlate of GRSS based on 41 genes (NGSS1). We also built an alternate model based upon 13 genes associated with severe infection acutely but displaying stable expression over time (NGSS2).ResultsNGSS1 is strongly correlated with the disease severity, demonstrating a naïve correlation (ρ) of ρ = 0.935 and cross-validated correlation of 0.813. As a binary classifier (mild versus severe), NGSS1 correctly classifies disease severity in 89.6% of the subjects following cross-validation. NGSS2 has slightly less, but comparable, accuracy with a cross-validated correlation of 0.741 and classification accuracy of 84.0%.ConclusionAirway gene expression patterns, obtained following a minimally-invasive procedure, have potential utility for development of clinically useful biomarkers that correlate with disease severity in primary RSV infection.

Prospective randomized trial of gene expression classifier utility following radical prostatectomy (G-MINOR).

15 Background: Decipher is a tissue-based genomic classifier (GC) developed and validated in the post-radical prostatectomy (RP) setting as a predictor of metastasis. We conducted the first prospective randomized controlled trial assessing the use of a prostate cancer GC, with a primary objective to determine the impact of test results on adjuvant treatment decisions. Methods: The Genomics in Michigan ImpactiNg Observation or Radiation (G-MINOR) randomized trial enrolled participants across 12 centers between January 2017-August 2018. Eligible patients had undergone RP within 9 months of enrollment, had pT3-4 disease and/or positive surgical margins, and a PSA &lt; 0.1ng/mL. Patients were assigned to either the GC or Usual Care (UC) group using cluster-crossover block randomization. Patients and providers in both arms received a CAPRA-S recurrence risk score. Decipher scores were obtained on RP tissue of all patients, but patients and providers in the UC arm were blinded to the results. The primary endpoint was the impact of impact of GC test result on adjuvant treatment decisions compared to clinical factors alone within 18 months of RP. Results: 356 patients were randomized and 340 had at least 18 months of follow-up. Of these, all but 2 control (UC) patients had sufficient tissue to pass quality control for GC testing. Randomization resulted in 175 (51.5%) GC and 165 (48.5%) UC patients. There were no significant differences in clinical variables or Decipher scores between arms. At 18 months post-RP, 19 (10.9%) patients in the GC group and 12 (7.3%) patients in the UC group had received adjuvant treatment. In the primary analysis, availability of the GC score in the GC arm was significantly associated with adjuvant treatment in GC high-risk patients after controlling for CAPRA-S risk (OR 7.6, 95%CI 1.95-29.6, p = 0.009). In the GC arm, both GC score (OR 8.8, 95%CI 1.9-39.7, p = 0.005) and CAPRA-S score (OR 3.8, 95%CI 1.09-12.9, p = 0.04) were independently associated with adjuvant treatment in a multivariable logistic regression model. Conclusions: In the first ever randomized trial testing the impact of a prostate cancer genomic classifier on treatment decisions, the use of a GC post-RP impacted post-operative treatment in a manner concordant with classifier risk. Further follow-up will be necessary to assess the impact of GC testing on oncologic outcomes. Clinical trial information: NCT02783950. [Table: see text]

Integrated Analysis of Methylomic and Transcriptomic Data to Identify Potential Diagnostic Biomarkers for Major Depressive Disorder.

Major depressive disorder (MDD) is a mental illness with high incidence and complex etiology, that poses a serious threat to human health and increases the socioeconomic burden. Currently, high-accuracy biomarkers for MDD diagnosis are urgently needed. This paper aims to identify novel blood-based diagnostic biomarkers for MDD. Whole blood DNA methylation data and gene expression data from the Gene Expression Omnibus database are downloaded. Then, differentially expressed/methylated genes (DEGs/DMGs) are identified. In addition, we made a systematic analysis of the DNA methylation on 5′-C-phosphate-G-3′ (CpGs) in all of the gene regions, as well as different gene regions, and then we defined a “dominant” region. Subsequently, integrated analysis is employed to identify the robust MDD-related blood biomarkers. Finally, a gene expression classifier and a methylation classifier are constructed using the random forest algorithm and the leave-one-out cross-validation method. Our results demonstrate that DEGs are mainly involved in the inflammatory response-associated pathways, while DMGs are primarily concentrated in the neurodevelopment- and neuroplasticity-associated pathways. Our integrated analysis identified 46 hypo-methylated and up-regulated (hypo-up) genes and 71 hyper-methylated and down-regulated (hyper-down) genes. One gene expression classifier and two DNA methylation classifiers, based on the CpGs in all of the regions or in the dominant regions are constructed. The gene expression classifier possessed the best predictive ability, followed by the DNA methylation classifiers, based on the CpGs in both the dominant regions and all of the regions. In summary, the integrated analysis of DNA methylation and gene expression has identified 46 hypo-up genes and 71 hyper-down genes, which could be used as diagnostic biomarkers for MDD.

Prevalence of cancer and the benign call rate of afirma gene classifier in 18 F-Fluorodeoxyglucose positron emission tomography positive cytologically indeterminate thyroid nodules.

Background 18 F‐Fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) positive (PET+) cytologically indeterminate thyroid nodules (ITNs) have variable cancer risk in the literature. The benign call rate (BCR) of Afirma Gene Classifier (Gene Expression Classifier, GEC, or Genome Sequence Classifier, GSC) in (PET +) ITNs is unknown.MethodsThis is a retrospective study at our institution of all patients with (PET+) ITNs (Bethesda III/IV) from 1 January 2010 to 21 May 2019 who underwent Afirma testing and/or surgery or repeat FNA with benign cytology.ResultsForty‐five (PET+) ITNs were identified: 31 Afirma‐tested (GEC = 20, GSC = 11) and 14 either underwent surgery (n = 13) or repeat FNA (Benign cytology) (n = 1) without Afirma. The prevalence of cancer and noninvasive follicular thyroid neoplasm with papillary‐like nuclear features (NIFTP) including only resected nodules and ITN with repeat benign FNA (n = 33) was 36.4% (12/33). Excluding all Afirma “suspicious” non‐resected ITNs and assuming all Afirma “benign” ITNs were truly benign, that prevalence was 28.6% (12/42). The BCR with GSC was 64% compared to 25% with GEC (p = 0.056). Combining GSC/GEC‐tested ITNs, the BCR was higher in ITNs demonstrating low/very low‐risk sonographic pattern by the American Thyroid Association (ATA) classification and ITNs scoring <4 by the American College of Radiology Thyroid Imaging, Reporting and Data System (ACR‐TI‐RADS) than ITNs with higher sonographic pattern/score (p = 0.025).ConclusionsThe prevalence of cancer/NIFTP in (PET+) ITNs was 28.6–36.4% depending on the method of calculation. The BCR of Afirma GSC was 64%. Combining Afirma GEC/GSC‐tested ITNs, BCR was higher in ITNs with a lower risk sonographic pattern.

Detection of Chemotherapy-resistant Pancreatic Cancer Using a Glycan Biomarker, sTRA.

A subset of pancreatic ductal adenocarcinomas (PDACs) is highly resistant to systemic chemotherapy, but no markers are available in clinical settings to identify this subset. We hypothesized that a glycan biomarker for PDACs called sialylated tumor-related antigen (sTRA) could be used for this purpose. We tested for differences between PDACs classified by glycan expression in multiple systems: sets of cell lines, organoids, and isogenic cell lines; primary tumors; and blood plasma from human subjects. The sTRA-expressing models tended to have stem-like gene expression and the capacity for mesenchymal differentiation, in contrast to the nonexpressing models. The sTRA cell lines also had significantly increased resistance to seven different chemotherapeutics commonly used against pancreatic cancer. Patients with primary tumors that were positive for a gene expression classifier for sTRA received no statistically significant benefit from adjuvant chemotherapy, in contrast to those negative for the signature. In another cohort, based on direct measurements of sTRA in tissue microarrays, the patients who were high in sTRA again had no statistically significant benefit from adjuvant chemotherapy. Furthermore, a blood plasma test for the sTRA glycan identified the PDACs that showed rapid relapse following neoadjuvant chemotherapy. This research demonstrates that a glycan biomarker could have value to detect chemotherapy-resistant PDAC in clinical settings. This capability could aid in the development of stratified treatment plans and facilitate biomarker-guided trials targeting resistant PDAC.

The Cancer Senescopedia – Delineation of Cancer Cell Senescence

Cellular senescence is characterized as a stable proliferation arrest that can be triggered by multiple stresses. Most knowledge about senescent cells has been obtained from studies in primary cells. However, senescence features may be different in cancer cells, since the pathways that are involved in senescence induction are often deregulated in cancer. We report here a comprehensive analysis of the transcriptome and senolytic responses in a panel of 13 cancer cell lines rendered senescent by two distinct compounds. We show that in cancer cells the response to senolytic agents and the composition of the senescence-associated secretory phenotype are determined mainly by cell line intrinsic properties, rather than the senescence trigger. Using machine learning, we established the SENCAN gene expression classifier that can be used for the detection of senescence in cancer cell samples. The expression profiles and senescence classifier are available as an interactive online Cancer SENESCopedia.

1216. Benchmarking Published Gene Signatures for Robust Infection Classification

BackgroundHost gene expression has emerged as a promising diagnostic strategy to discriminate bacterial and viral infection. Multiple gene signatures of varying size and complexity have been developed in various clinical populations. However, there has been no systematic comparison of these signatures. It is also unclear how these signatures apply to different clinical populations. This meta-analysis examined 19 published signatures, validated in 49 publicly available datasets for a total of 4750 patients. The objectives were to understand how the signatures compared to each other with respect to composition and performance, and to evaluate their performance in different patient subgroups.MethodsSignatures were characterized with respect to size, platform, and discovery population. For each of the 19 signatures, we ran leave-one-out cross-validation to generate AUCs for bacterial classification and viral classification. We then applied dataset-specific thresholds to generate accuracies, pooling patients across datasets.ResultsSignature performance varied significantly with a median AUC across all validation datasets ranging from 0.55 to 0.94 for bacterial classification and 0.79 to 0.96 for viral classification. Signature size varied (1- 341 genes) with smaller signatures generally performing more poorly for both bacterial classification (P < .001) and for viral classification (P = 0.02). Viral infection was easier to diagnose than bacterial infection (85% vs. 80% overall accuracy, respectively; P < .001). Host gene expression classifiers performed more poorly in children < 12-years compared to those older than 12-years for both bacterial infection (77% vs. 83%, respectively; P < .001) and for viral infection (82% vs. 89%, respectively; P < .001). We did not observe differences based on illness severity as defined by ICU care for either bacterial or viral infections.ConclusionWe observed significant differences among gene expression signatures for bacterial/viral discrimination, though these were not due to variations in the discovery methods or populations. Signature size directly correlated with test performance, which was generally better for the diagnosis of viral infection and in populations >12-years.Disclosures Ephraim L. Tsalik, MD, MHS, PhD, Predigen (Shareholder, Other Financial or Material Support, Founder)

Tumor infiltrating neutrophils and gland formation predict overall survival and molecular subgroups in pancreatic ductal adenocarcinoma.

ABSTRACTBackgroundRNA‐sequencing‐based classifiers can stratify pancreatic ductal adenocarcinoma (PDAC) into prognostically significant subgroups but are not practical for use in clinical workflows. Here, we assess whether histomorphological features may be used as surrogate markers for predicting molecular subgroup and overall survival in PDAC.MethodsNinety‐six tissue samples from 50 patients with non‐resectable PDAC were scored for gland formation, stromal maturity, mucin, necrosis, and neutrophil infiltration. Prognostic PDAC gene expression classifiers were run on all tumors using whole transcriptome sequencing data from the POG trial (NCT02155621). Findings were validated using digital TCGA slides (n = 50). Survival analysis used multivariate Cox proportional‐hazards tests and log‐rank tests.ResultsThe combination of low gland formation and low neutrophil infiltration was significantly associated with the poor prognosis PDAC molecular subgroup (basal‐like or squamous) and was an independent predictor of shorter overall survival, in both frozen section (n = 47) and formalin‐fixed paraffin‐embedded (n = 49) tissue samples from POG patients, and in the TCGA samples. This finding held true in the subgroup analysis of primary (n = 17) and metastatic samples (n = 79). The combination of high gland formation and high neutrophils had low sensitivity but high specificity for favorable prognosis subgroups.ConclusionsThe assessment of gland formation and neutrophil infiltration on routine histological sections can aid in prognostication and allow inferences to be made about molecular subtype, which may help guide patient management decisions and contribute to our understanding of heterogeneity in treatment response.