Independent Validation Cohort Research Articles

High-throughput proteomics of breast cancer subtypes: Biological characterization and multiple candidate biomarker panels to patients' stratification

Background and aimsThe actual classification of breast tumors in subtypes represents an attempt to stratify patients into clinically cohesive groups, nevertheless, clinicians still lack reproducible and reliable protein biomarkers for breast cancer subtype discrimination. In this study, we aimed to access the differentially expressed proteins between these tumors and its biological implications, contributing to the subtype's biological and clinical characterization, and with protein panels for subtype discrimination. MethodsIn our study, we applied high-throughput mass spectrometry, bioinformatic, and machine learning approaches to investigate the proteome of different breast cancer subtypes. ResultsWe identified that each subtype depends on different protein expression patterns to sustain its malignancy, and also alterations in pathways and processes that can be associated with each subtype and its biological and clinical behaviors. Regarding subtype biomarkers, our panels achieved performances with at least 75% of sensibility and 92% of specificity. In the validation cohort, the panels obtained acceptable to outstanding performances (AUC = 0.740 to 1.00). ConclusionsIn general, our results expand the accuracy of breast cancer subtypes' proteomic landscape and improve the understanding of its biological heterogeneity. In addition, we identified potential protein biomarkers for the stratification of breast cancer patients, improving the repertoire of reliable protein biomarkers. SignificanceBreast cancer is the most diagnosed cancer type worldwide and the most lethal cancer in women. As a heterogeneous disease, breast cancer tumors can be classified into four major subtypes, each presenting particular molecular alterations, clinical behaviors, and treatment responses. Thus, a pivotal step in patient management and clinical decisions is accurately classifying breast tumor subtypes. Currently, this classification is made by the immunohistochemical detection of four classical markers (estrogen receptor, progesterone receptor, HER2 receptor, and the Ki-67 index); however, it is known that these markers alone do not fully discriminate the breast tumor subtypes. Also, the poor understanding of the molecular alterations of each subtype leads to a challenging decision-making process regarding treatment choice and prognostic determination. This study, through high-throughput label-free mass-spectrometry data acquisition and downstream bioinformatic analysis, advances in the proteomic discrimination of breast tumors and achieves an in-depth characterization of the subtype's proteomes. Here, we indicate how the variations in the subtype's proteome can influence the tumor's biological and clinical differences, highlighting the variation in the expression pattern of oncoproteins and tumor suppressor proteins between subtypes. Also, through our machine-learning approach, we propose multi-protein panels with the potential to discriminate the breast cancer subtypes. Our panels achieved high classification performance in our cohort and in the independent validation cohort, demonstrating their potential to improve the current tumor discrimination system as complements to the classical immunohistochemical classification.

Plasma lipidomic profiling of thiopurine-induced leukopenia after NUDT15 genotype-guided dosing in Chinese IBD patients.

Thiopurines, azathiopurine (AZA) and mercaptopurine (6-MP) have been regularly used in the treatment of inflammatory bowel disease (IBD). Despite optimized dosage adjustment based on the NUDT15 genotypes, some patients still discontinue or change treatment regimens due to thiopurine-induced leukopenia. We proposed a prospective observational study of lipidomics to reveal the lipids perturbations associated with thiopurine-induced leukopenia. One hundred and twenty-seven IBD participants treated with thiopurine were enrolled, twenty-seven of which have developed thiopurine-induced leucopenia. Plasma lipid profiles were measured using Ultra-High-Performance Liquid Chromatography-Tandem Q-Exactive. Lipidomic alterations were validated with an independent validation cohort (leukopenia n = 26, non-leukopenia n = 74). Using univariate and multivariate analysis, there were 16 lipid species from four lipid classes, triglyceride (n = 11), sphingomyelin (n = 1), phosphatidylcholine (n = 1) and lactosylceramide (n = 3) identified. Based on machine learning feature reduction and variable screening strategies, the random forest algorithm established by six lipids showed an excellent performance to distinguish the leukopenia group from the normal group, with a model accuracy of 95.28% (discovery cohort), 79.00% (validation cohort) and an area under the receiver operating characteristic (ROC) curve (ROC-AUC) of 0.9989 (discovery cohort), 0.8098 (validation cohort). Our novel findings suggested that lipidomic provided unique insights into formulating individualized medication strategies for thiopurines in IBD patients.

Exome Sequencing Implicates DGKZ , ESRRA , and GXYLT1 for Modulating Granuloma Formation in Crohn Disease.

Non-caseating granulomas may indicate a more aggressive phenotype of Crohn disease (CD). Genetic associations of granulomatous CD (GCD) may help elucidate disease pathogenesis. Whole-exome sequencing was performed on peripheral blood-derived DNA from 17 pediatric patients with GCD and 19 with non-GCD (NGCD), and from an independent validation cohort of 44 GCD and 19 NGCD cases. PLINK (a tool set for whole-genome association and population-based linkage analyses) analysis was used to identify single nucleotide polymorphisms (SNPs) differentiating between groups, and subgroup allele frequencies were also compared to a public genomic database (gnomAD). The Combined Annotation Dependent Depletion scoring tool was used to predict deleteriousness of SNPs. Human leukocyte antigen (HLA) haplotype findings were compared to a control group (n = 8496). PLINK-based analysis between GCD and NGCD groups did not find consistently significant hits. gnomAD control comparisons, however, showed consistent subgroup associations with DGKZ , ESRRA , and GXYLT1 , genes that have been implicated in mammalian granulomatous inflammation. Our findings may guide future research and precision medicine.

Proteome Landscapes of Human Hepatocellular Carcinoma and Intrahepatic Cholangiocarcinoma

Liver cancer is among the top leading causes of cancer mortality worldwide. Particularly, hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (CCA) have been extensively investigated from the aspect of tumor biology. However, a comprehensive and systematic understanding of the molecular characteristics of HCC and CCA remains absent. Here, we characterized the proteome landscapes of HCC and CCA using the data-independent acquisition (DIA) mass spectrometry (MS) method. By comparing the quantitative proteomes of HCC and CCA, we found several differences between the two cancer types. In particular, we found an abnormal lipid metabolism in HCC and activated extracellular matrix-related pathways in CCA. We next developed a three-protein classifier to distinguish CCA from HCC, achieving an area under the curve (AUC) of 0.92, and an accuracy of 90% in an independent validation cohort of 51 patients. The distinct molecular characteristics of HCC and CCA presented in this study provide new insights into the tumor biology of these two major important primary liver cancers. Our findings may help develop more efficient diagnostic approaches and new targeted drug treatments.

Delineating COVID-19 subgroups using routine clinical data identifies distinct in-hospital outcomes

The COVID-19 pandemic has been a great challenge to healthcare systems worldwide. It highlighted the need for robust predictive models which can be readily deployed to uncover heterogeneities in disease course, aid decision-making and prioritise treatment. We adapted an unsupervised data-driven model—SuStaIn, to be utilised for short-term infectious disease like COVID-19, based on 11 commonly recorded clinical measures. We used 1344 patients from the National COVID-19 Chest Imaging Database (NCCID), hospitalised for RT-PCR confirmed COVID-19 disease, splitting them equally into a training and an independent validation cohort. We discovered three COVID-19 subtypes (General Haemodynamic, Renal and Immunological) and introduced disease severity stages, both of which were predictive of distinct risks of in-hospital mortality or escalation of treatment, when analysed using Cox Proportional Hazards models. A low-risk Normal-appearing subtype was also discovered. The model and our full pipeline are available online and can be adapted for future outbreaks of COVID-19 or other infectious disease.

Fragmentomic analysis of circulating tumor DNA-targeted cancer panels

The isolation of cell-free DNA (cfDNA) from the bloodstream can be used to detect and analyze somatic alterations in circulating tumor DNA (ctDNA), and multiple cfDNA-targeted sequencing panels are now commercially available for Food and Drug Administration (FDA)-approved biomarker indications to guide treatment. More recently, cfDNA fragmentation patterns have emerged as a tool to infer epigenomic and transcriptomic information. However, most of these analyses used whole-genome sequencing, which is insufficient to identify FDA-approved biomarker indications in a cost-effective manner. We used machine learning models of fragmentation patterns at the first coding exon in standard targeted cancer gene cfDNA sequencing panels to distinguish between cancer and non-cancer patients, as well as the specific tumor type and subtype. We assessed this approach in two independent cohorts: a published cohort from GRAIL (breast, lung, and prostate cancers, non-cancer, n = 198) and an institutional cohort from the University of Wisconsin (UW; breast, lung, prostate, bladder cancers, n= 320). Each cohort was split 70%/30% into training and validation sets. In the UW cohort, training cross-validated accuracy was 82.1%, and accuracy in the independent validation cohort was 86.6% despite a median ctDNA fraction of only 0.06. In the GRAIL cohort, to assess how this approach performs in very low ctDNA fractions, training and independent validation were split based on ctDNA fraction. Training cross-validated accuracy was 80.6%, and accuracy in the independent validation cohort was 76.3%. In the validation cohort where the ctDNA fractions were all <0.05 and as low as 0.0003, the cancer versus non-cancer area under the curve was 0.99. To our knowledge, this is the first study to demonstrate that sequencing from targeted cfDNA panels can be utilized to analyze fragmentation patterns to classify cancer types, dramatically expanding the potential capabilities of existing clinically used panels at minimal additional cost.

Combination of alpha-fetoprotein and neutrophil-to-lymphocyte ratio to predict treatment response and survival outcomes of patients with unresectable hepatocellular carcinoma treated with immune checkpoint inhibitors

BackgroundImmune-checkpoint inhibitors (ICIs) have revolutionized the treatment of hepatocellular carcinoma (HCC). However, long-term survival outcomes and treatment response of HCC patients undergoing immunotherapy is unpredictable. The study aimed to evaluate the role of alpha-fetoprotein (AFP) combined with neutrophil-to-lymphocyte ratio (NLR) to predict the prognosis and treatment response of HCC patients receiving ICIs.MethodsPatients with unresectable HCC who received ICI treatment were included. The HCC immunotherapy score was developed from a retrospective cohort at the Eastern Hepatobiliary Surgery Hospital to form the training cohort. The clinical variables independently associated with overall survival (OS) were identified using univariate and multivariate Cox regression analysis. Based on multivariate analysis of OS, a predictive score based on AFP and NLR was constructed, and patients were stratified into three risk groups according to this score. The clinical utility of this score to predict progression-free survival (PFS) and differentiate objective response rate (ORR) and disease control rate (DCR) was also performed. This score was validated in an independent external validation cohort at the First Affiliated Hospital of Wenzhou Medical University.ResultsBaseline AFP ≤ 400 ng/ml (hazard ratio [HR] 0.48; 95% CI, 0.24–0.97; P = 0.039) and NLR ≤ 2.77 (HR 0.11; 95% CI, 0.03–0.37; P<0.001) were found to be independent risk factors of OS. The two labolatory values were used to develop the score to predict survival outcomes and treatment response in HCC patients receiving immunotherapy, which assigned 1 point for AFP > 400 ng/ml and 3 points for NLR > 2.77. Patients with 0 point were classified as the low-risk group. Patients with 1–3 points were categorized as the intermediate-risk group. Patients with 4 points were classified as the high-risk group. In the training cohort, the median OS of the low-risk group was not reached. The median OS of the intermediate-risk group and high-risk group were 29.0 (95% CI 20.8–37.3) months and 16.0 (95% CI 10.8–21.2) months, respectively (P < 0.001). The median PFS of the low-risk group was not reached. The median PFS of the intermediate-risk group and high-risk group were 14.6 (95% CI 11.3–17.8) months and 7.6 (95% CI 3.6–11.7) months, respectively (P < 0.001). The ORR and DCR were highest in the low-risk group, followed by the intermediate-risk group and the high-risk group (P < 0.001, P = 0.007, respectively). This score also had good predictive power using the validation cohort.ConclusionThe HCC immunotherapy score based on AFP and NLR can predict survival outcomes and treatment response in patients receiving ICI treatments, suggesting that this score could serve as a useful tool for identification of HCC patients likely to benefit from immunotherapy.

#4835 KIDNEY FUNCTION-SPECIFIC CUT-OFF VALUES OF HIGH-SENSITIVITY CARDIAC TROPONIN T FOR THE DIAGNOSIS OF ACUTE MYOCARDIAL INFARCTION

Abstract Background and Aims Acute myocardial infarction (AMI) is the leading cause of death worldwide. Patients with kidney dysfunction have high prevalence of AMI, thus early detection of AMI is necessary. However, in patients with decreased estimated glomerular filtration rate (eGFR), using the traditional cut-off value (14 ng/L) of high-sensitivity cardiac troponin T (hs-cTnT) resulted in lower specificity for diagnosing AMI. Our study aims to determine and validate kidney function-specific cut-off values of hs-cTnT for diagnosing AMI in patients with different degrees of kidney dysfunction. Method This large, multicenter, observational cohort study included 21,912 hospitalized patients who had undergone coronary angiography in 6 clinical centers from 2013 to 2021. The study outcome was the diagnosis of AMI. In the derivation cohort, kidney function-specific cut-off values of hs-cTnT for diagnosing AMI were determined to improve the specificity without the cost of sensitivity, as compared to that using 14 ng/L as the cut-off value in the normal kidney function group. The efficacy of the novel cut-off values was validated in an independent validation cohort enrolled from another clinical center. Results In the derivation cohort (n = 12,900), 3,247 patients (25.2%) had an eGFR &amp;lt;60 mL/min/1.73m2. Even in the absence of AMI, 1485 (50.2%) participants with eGFR &amp;lt;60 mL/min/1.73m2 had a hs-cTnT concentration ≥14 ng/L. Using 14 ng/L as the threshold of hs-cTnT for diagnosing AMI led to a significantly reduced specificity (9.1–52.7% vs. 75.0%) and positive predictive value (15.5–21.1% vs. 28.7%) in patients with kidney dysfunction, as compared with patients with normal kidney function. The kidney function-specific cut-off values (eGFR ≥60 ml/min/1.73m2: 14 ng/L, 60&amp;gt;eGFR≥30 mL/min/1.73m2: 18 ng/L, and eGFR &amp;lt;30 ml/min/1.73m2: 48 ng/L) were determined and remarkably improved the diagnostic accuracy for AMI in participants with different levels of kidney dysfunction (specificity: reach to 52.8–63.0%; positive predictive value: 18.7–33.6%), without compromising sensitivity (96.6–97.9%). The kidney function-specific cut-off values increased the diagnosis accuracy of hs-cTnT for diagnosing AMI in every eGFR subgroup in the independent validation cohort (n = 8,012). Conclusion The kidney function-specific cut-off values of hs-cTnT significantly improved the diagnostic accuracy of AMI in patients with kidney dysfunction, and might be generally useful in clinical practice.

Prognostic impact of intra- and peritumoral immune cell subpopulations in head and neck squamous cell carcinomas - comprehensive analysis of the TCGA-HNSC cohort and immunohistochemical validation on 101 patients.

Due to the expanding role of immune checkpoint inhibition in the treatment of head and neck squamous cell carcinoma, understanding immunological processes in the tumor microevironment (TME) has strong translational importance. Though analytical methods for a comprehensive analysis of the immunological TME have constantly improved and expanded over the past years the prognostic relevance of immune cell composition in head and neck cancer TME largely remains ambiguous with most studies focusing on one or a small subset of immune cells. The overall survival (OS) of the TCGA-HNSC patient cohort comprising 513 head and neck cancer patients was correlated with a total of 29 different immune metrics including a wide spectrum of immune cell subpopulations as well as immune checkpoint receptors and cytokines using RNAseq based immune deconvolution analyses. The most significant predictors of survival among these 29 immune metrics were validated on a separate HNSCC patient cohort (n=101) using immunohistochemistry: CD3, CD20+CXCR5, CD4+CXCR5, Foxp3 and CD68. Overall immune infiltration irrespective of immune cell composition showed no significant correlation with the patients' overall survival in the TCGA-HNSC cohort. However, when focusing on different immune cell subpopulations, naïve B cells (p=0.0006), follicular T-helper cells (p<0.0001), macrophages (p=0.0042), regulatory T cells (p=0.0306), lymphocytes (p=0.0001), and cytotoxic T cells (p=0.0242) were identified as highly significant predictors of improved patient survival. Using immunohistochemical detection of these immune cells in a second independent validation cohort of 101 HNSCC patients, we confirmed the prognostic relevance of follicular T helper cells, cytotoxic T cells and lymphocytes. In multivariable analysis, HPV negativity and advanced UICC stages were identified as additional prognostic biomarkers associated with poor outcome. Our study highlights the prognostic relevance of the immunological tumor environment in head and neck cancer and demonstrates that a more detailed analysis of immune cell composition and immune cell subtypes is necessary to accurately prognosticate. We observed the highest prognostic relevance for lymphocytes, cytotoxic T cells, and follicular T helper cells, suggesting further investigations focusing on these specific immune cell subpopulations not only as predictors of patient prognosis but also as promising targets of new immunotherapeutic strategies.

Identification of BRCA1/2 mutation female carriers using circulating microRNA profiles

Identifying germline BRCA1/2 mutation carriers is vital for reducing their risk of breast and ovarian cancer. To derive a serum miRNA-based diagnostic test we used samples from 653 healthy women from six international cohorts, including 350 (53.6%) with BRCA1/2 mutations and 303 (46.4%) BRCA1/2 wild-type. All individuals were cancer-free before and at least 12 months after sampling. RNA-sequencing followed by differential expression analysis identified 19 miRNAs significantly associated with BRCA mutations, 10 of which were ultimately used for classification: hsa-miR-20b-5p, hsa-miR-19b-3p, hsa-let-7b-5p, hsa-miR-320b, hsa-miR-139-3p, hsa-miR-30d-5p, hsa-miR-17-5p, hsa-miR-182-5p, hsa-miR-421, hsa-miR-375-3p. The final logistic regression model achieved area under the receiver operating characteristic curve 0.89 (95% CI: 0.87–0.93), 93.88% sensitivity and 80.72% specificity in an independent validation cohort. Mutated gene, menopausal status or having preemptive oophorectomy did not affect classification performance. Circulating microRNAs may be used to identify BRCA1/2 mutations in patients of high risk of cancer, offering an opportunity to reduce screening costs.

The diagnostic value of serum insulin-like growth factor binding protein 7 in gastric cancer.

Early detection might help in reducing the burden and promoting the survival rate of gastric cancers. Herein, we tried to explore the diagnostic value of insulin-like growth factor binding protein 7 (IGFBP7) in gastric cancers. In this study, we first analyzed the expression levels and prognostic value of IGFBP7 mRNA in gastric cancers from The Cancer Genome Atlas (TCGA) database. Then, we recruited 169 gastric cancer patients and 100 normal controls as training cohort, and 55 gastric cancer patients and 55 normal controls as independent validation cohort. Enzyme-linked immunosorbent assay was applied to test the serum levels of IGFBP7. The receiver operating characteristic curve (ROC) and the area under the curve (AUC) were applied to evaluation the diagnostic value. TCGA showed that IGFBP7 mRNA was dysregulated and associated with prognosis in gastric cancer patients. Then, we examined the expression of serum IGFBP7 and found that serum IGFBP7 expressed lower in gastric cancer patients than normal controls both in training and independent validation cohorts (p < 0.0001). In training cohort, with the cutoff value of 1.515 ng/ml, the AUC for distinguishing gastric cancer patients was 0.774 (95% CI [0.713-0.836]) with sensitivity of 36.7% (95% CI [29.5-44.5]) and specificity of 90.0% (95% CI [82.0-94.8]). As for early-stage EJA, the AUC was 0.773 (95% CI [0.701-0.845]) with the sensitivity of 33.3% (95% CI [14.4-58.8]). In independent validation cohort, with the same cutoff value, the AUC reached to 0.758 (95% CI [0.664-0.852]). Similarly, for early-stage gastric cancer diagnosis in the independent validation cohort, the AUC value was 0.778 (95% CI [0.673-0.882]). This study indicated that serum IGFBP7 might act as a potential early diagnostic marker for gastric cancers.

Plasma amyloid beta 42 is a biomarker for patients with hereditary, but not sporadic, cerebral amyloid angiopathy

BackgroundThe diagnosis of probable cerebral amyloid angiopathy (CAA) is currently mostly based on characteristics of brain MRI. Blood biomarkers would be a cost-effective, easily accessible diagnostic method that may complement diagnosis by MRI and aid in monitoring disease progression. We studied the diagnostic potential of plasma Aβ38, Aβ40, and Aβ42 in patients with hereditary Dutch-type CAA (D-CAA) and sporadic CAA (sCAA).MethodsAll Aβ peptides were quantified in the plasma by immunoassays in a discovery cohort (11 patients with presymptomatic D-CAA and 24 patients with symptomatic D-CAA, and 16 and 24 matched controls, respectively) and an independent validation cohort (54 patients with D-CAA, 26 presymptomatic and 28 symptomatic, and 39 and 46 matched controls, respectively). In addition, peptides were quantified in the plasma in a group of 61 patients with sCAA and 42 matched controls. We compared Aβ peptide levels between patients and controls using linear regression adjusting for age and sex.ResultsIn the discovery cohort, we found significantly decreased levels of all Aβ peptides in patients with presymptomatic D-CAA (Aβ38: p < 0.001; Aβ40: p = 0.009; Aβ42: p < 0.001) and patients with symptomatic D-CAA (Aβ38: p < 0.001; Aβ40: p = 0.01; Aβ42: p < 0.001) compared with controls. In contrast, in the validation cohort, plasma Aβ38, Aβ40, and Aβ42 were similar in patients with presymptomatic D-CAA and controls (Aβ38: p = 0.18; Aβ40: p = 0.28; Aβ42: p = 0.63). In patients with symptomatic D-CAA and controls, plasma Aβ38 and Aβ40 were similar (Aβ38: p = 0.14; Aβ40: p = 0.38), whereas plasma Aβ42 was significantly decreased in patients with symptomatic D-CAA (p = 0.033). Plasma Aβ38, Aβ40, and Aβ42 levels were similar in patients with sCAA and controls (Aβ38: p = 0.092; Aβ40: p = 0.64. Aβ42: p = 0.68).ConclusionsPlasma Aβ42 levels, but not plasma Aβ38 and Aβ40, may be used as a biomarker for patients with symptomatic D-CAA. In contrast, plasma Aβ38, Aβ40, and Aβ42 levels do not appear to be applicable as a biomarker in patients with sCAA.

Abstract B025: MicroRNA regulators of neuroendocrine differentiation of prostate cancer

Abstract Background: Therapy-induced neuroendocrine prostate cancer (NEPC) is an extremely aggressive variant of castration-resistant prostate cancer (CRPC) that is increasing in incidence with the widespread use of second generation of androgen receptor (AR)-pathway inhibitors (APIs) such as Enzalutamide (ENZ) and Abiraterone. This aggressive variant arises from CRPC-Adenocarcinomas (CRPC-Adeno) via a reversible trans-differentiation process, referred to as neuroendocrine differentiation (NED) wherein cells undergo a lineage switch and exhibit neuroendocrine (NE) features, characterized by expression of neuronal markers such as enolase 2 (ENO2), chromogranin A (CHGA) and synaptophysin (SYP). There is an urgent need of identifying novel molecular markers to assess emergence of NED in CRPC patients. We demonstrated that progression of advanced CRPC with adenocarcinoma characteristics (CRPC-Adeno) to therapy-induced, androgen-independent NE (CRPC-NE) states is associated with a characteristic set of miRNA alterations that promote plasticity of advanced prostate adenocarcinomas to NEPC (Bhagirath et al., Oncogene, 2020). Importantly, we could develop a ‘novel miRNA classifier’ to robustly stratify CRPC-NE tumors from CRPC-Adenocarcinomas. Here we further validate the classifier in independent clinical cohorts and deduce the optimal miRNA genes required for NEPC diagnosis. We further study the functional significance of miR-28-3p, a miRNA identified by sequencing, in NEPC. Methods: Human clinical samples with corresponding clinical information were procured from two independent sites of Prostate Cancer Biorepository Network (PCBN). Samples included human CRPC-adeno vs CRPC-NE. FFPE sections from these clinical samples were microdissected, RNA were extracted and small RNA sequencing was performed using an Illumina NextSeq 500 platform. Sequencing data were analyzed and machine learning algorithms were applied (random forest machine learning technique with leave-pair-out cross validation (LPOCV)). The performance of classifier was measured using receiver operating characteristic (ROC) analysis with area under the curve (AUC) as the primary evaluation metric. The functional significance of miR-28-3p was studied by its overexpression and knockdown in PCa cell lines followed by functional assays. Results: Unsupervised analysis of sequencing data by principal component analyses (PCA) revealed distinct clustering of the CRPC-NE tumors from CRPC-Adenocarcinomas based on miRNA profiles suggesting that miRNA profiles can be used to stratify these tumor types. Our analyses showed that a set of 5 miRNAs of the classifier are important in distinguishing between CRPC-Adeno vs CRPC-NE with an AUC=0.8318. Further, miR 28-3p potentially has a biphasic role in PCa, where it is oncogenic in early stages and acts as a tumor suppressor in late stage PCa. Conclusions: A ‘5- miRNA’ classifier was validated to be of significance in two independent validation cohorts. We propose this miRNA classifier as an important tool for diagnosing NED in CRPC patients. Citation Format: Amritha Sreekumar, Sharanjot Saini. MicroRNA regulators of neuroendocrine differentiation of prostate cancer [abstract]. In: Proceedings of the AACR Special Conference: Advances in Prostate Cancer Research; 2023 Mar 15-18; Denver, Colorado. Philadelphia (PA): AACR; Cancer Res 2023;83(11 Suppl):Abstract nr B025.

Validation of the PDACai signature in predicting relative benefit from frontline FOLFIRINOX (FFX) and gemcitabine/nab-paclitaxel (GA) for patients (pts) with metastatic pancreatic cancer (mPDAC).

4149 Background: Nearly 50% of pts with mPDAC never receive a 2nd line of therapy for metastatic disease following frontline FFX or GA. Genomic alterations in the DDR pathway (e.g. BRCA1/2) are associated with increased progression-free survival (PFS) on platinum-containing regimens (e.g. FFX), but other biomarker-treatment associations that predict benefit from GA and/or FFX in mPDAC remain unexplored. In this retrospective real-world evidence (RWE) study, we use a data-driven machine learning approach to gain new insights from the mutational landscape in mPDAC and validate the PDACai signature in predicting personalized benefit from both FFX and GA. Methods: We analyzed real-world outcomes from 711 pts with mPDAC who underwent genomic profiling via the Know Your Tumor program or were referred to Perthera by treating oncologists. Chart-abstracted PFS data on either 1st line FFX or GA were split (60:40) into independent training and validation cohorts for each regimen. All models integrate a shared set of clinical (age &lt; 63, sex) and lab-agnostic molecular features derived from clinical NGS testing reports (DDR pathway alterations, specific KRAS variants, frequently mutated genes). Relative benefit scores predicted by FFX or GA models were evenly binned into three PDACai signature categories representing lower, middle, and upper tertiles for each independent cohort. Statistical differences in median PFS were evaluated using ordinal Cox regression. Results: Median PFS followed predicted trends as generated by PDACai for each therapy in training and validation cohorts. The predictive utility of PDACai was confirmed in the independent validation cohorts when comparing PFS on FFX (p = 0.03744; HR = 0.75 [95% CI: 0.58-0.98]) and GA (p = 0.00006861; HR = 0.65 [95% CI: 0.53-0.8]) across tertiles. Conclusions: Response to chemotherapy is heterogeneous and difficult to predict in pts with mPDAC. Using RWE, the PDACai signature successfully predicted relative differences in PFS on both FFX and GA in mPDAC. With prospective validation, personalized insights generated by PDACai from pts with similar biomarker profiles could be used to tailor treatment sequencing in pts with NGS testing results, particularly those without actionable genomic findings. [Table: see text]

Detecting early stage breast cancer using low-depth cell-free DNA fragmentomics: A multi-center cohort study.

3075 Background: Breast cancer (BC) has contributed to the most cancer-related mortalities among females in 2020. Early detection using breast imaging methods, such as mammography and ultrasound, has significantly improved patients' survival. However, these early-screening methods suffer from high false-positive rates, resulting in many unnecessary biopsies which add to patients' discomfort. Cell-free DNA (cfDNA) fragmentomics assays have recently illustrated prominent abilities for detecting various cancer. In this multi-center prospective cohort study, we aim to develop a non-invasive liquid biopsy assay utilizing cfDNA fragmentomics and machine learning for detecting early-stage BC patients. Methods: We recruited 402 female patients (234 with BC and 168 with benign nodules [BN]) from the Yantai Yuhuangding Hospital (training cohort, N = 193, 91 BC and 102 BN) and the Cancer Hospital of the Chinese Academy of Medical Sciences (independent validation cohort, N = 209, 143 BC and 66 BN). Surgical or core needle biopsies were performed for all 402 patients following positive breast imaging results by mammography and ultrasound. Each patient's malignant or benign status was then pathologically confirmed using the tissue specimen. Women with negative biopsy results were excluded from breast cancer through a 6-month follow-up. The fragmentomics profiles, which were generated using plasma cfDNA WGS data (~5X), were used by our machine-learning algorithm. An automated machine learning approach was deployed to optimize base models before being ensembled into the final predictive model. Results: Our predictive model showed excellent performance in discriminating BCs from BNs, reaching an Area Under the Curve (AUC) of 0.816 and 0.811 in the training cohort (5-fold cross-validation) and the independent validation cohort, respectively. It achieved 74.1% sensitivity (95% CI: 66.1-81.1%) at 81.8% specificity (95% CI: 70.4-90.2%) in the validation cohort. In a potential clinical setting, our model could detect 53.0% (95% CI: 40.3-65.4%) false positives from the breast imaging methods at a targeted 90% sensitivity (89.5%, 95% CI:83.3–94.0%), which could theistically reduce more than half of the unnecessary biopsies. Moreover, our model illustrated great predictive power (AUC = 0.830) among small nodules (≤ 1cm), which were difficult to distinguish by the traditional imaging methods (AUC = 0.690). Finally, our model maintained its predictive power during a down-sampling process, even using 2X WGS data in the independent validation cohort (mean AUC: 0.789, 0.781-0.798). Conclusions: Our non-invasive liquid biopsy assay, which utilized low-depth cfDNA fragmentomics profiling and machine learning, can distinguish malignant nodules from benign nodules. The highlight of our model is its ability to prevent unnecessary biopsies among patients with false positive breast imaging results.

Spatially resolved multi-region transcriptomic subtyping and assessment of gene expression profiles associated with long-term benefit from chemo-immunotherapy in patients with extensive-stage small cell lung cancer (ES-SCLC).

8594 Background: Transcriptomic subtyping holds promise for personalized therapy in SCLC, but intratumoral transcriptomic heterogeneity and its clinical significance remain poorly defined. In this study, we aimed to assess transcription factor defined subtypes within multiple regions of intact tissues and identify gen sets associated with long-term chemo-immunotherapy benefit. Methods: We assessed baseline FFPE tumors from 32 ES-SCLC patients enrolled in the IMfirst clinical trial (EudraCT: 2019-002784-10). We used GeoMx DSP to perform transcriptomic analysis from multiple intratumoral regions of interest (ROIs). We used an assay with +1800 genes (GeoMx CTA) plus custom-designed mRNA probes targeting subtype-defining transcription factors not included in the CTA assay (POU2F3, NEUROD1, and YAP1). Custom probes were quantitatively validated using FFPE cell lines. Results: We profiled 175 ROIs from 32 tumors. Cluster analysis based on the expression of ASCL1, NEUROD1, POU2F3, and YAP1 showed that all samples clustered within 4 groups: SCLC-A (76 ROIs [43%]), SCLC-N (31 ROIs [18%]), SCLC-P (18 ROIs [10%]), and SCLC-Y (50 ROIs [29%]). ASCL1 was the most abundantly expressed transcription factor, prevailed in the SCLC-A subtype, and showed inverse correlation with POU2F3 (r=-0.55, p&lt;0.0001) and YAP1(r=-0.70, p&lt;0.0001). YAP1 was expressed at low levels and was more broadly distributed across all 4 subtypes. Differential expression and gene set enrichment analysis (GSEA) using linear mixed models revealed that SCLC-A subtype was enriched in cell cycle and DNA damage repair genes, and showed repression of multiple gene sets associated with anti-tumor immune response. In contrast, SCLC-Y subtype showed the opposite pattern. The SCLC-P subtype was also enriched in genes related to cancer antigens and T-cell checkpoints. Most patients (n=18, 56%) had tumors with coexistence of more than one subtype, not evident through morphological inspection. Combined SCLC-A and SCLC-Y subtypes was the most common mixed phenotype (n=8, 25%). Four patients (13%) had tumors with co-existence of three subtypes. Transcriptional subtypes, subtype-defining transcription factors as single genes, or the presence of subtype heterogeneity, were not associated with outcomes. Gene sets involved in mitochondrial metabolism and MHC class I antigen presentation were the highest enriched pathways in tumors from patients with sustained benefit (time to progression ³ 12 months). Conclusions: This study reveals substantial intratumoral transcriptomic subtype heterogeneity in human SCLC. In this limited sample set, we did not observe outcome association for transcriptional subtypes. Our findings related to long-term chemo-immunotherapy benefit require validation in independent cohorts.

Artificial intelligence-assisted ultrasound image analysis to discriminate early breast cancer in Chinese population: a retrospective, multicentre, cohort study.

Early diagnosis of breast cancer has always been a difficult clinical challenge. We developed a deep-learning model EDL-BC to discriminate early breast cancer with ultrasound (US) benign findings. This study aimed to investigate how the EDL-BC model could help radiologists improve the detection rate of early breast cancer while reducing misdiagnosis. In this retrospective, multicentre cohort study, we developed an ensemble deep learning model called EDL-BC based on deep convolutional neural networks. The EDL-BC model was trained and internally validated on B-mode and color Doppler US image of 7955 lesions from 6795 patients between January 1, 2015 and December 31, 2021 in the First Affiliated Hospital of Army Medical University (SW), Chongqing, China. The model was assessed by internal and external validations, and outperformed radiologists. The model performance was validated in two independent external validation cohorts included 448 lesions from 391 patients between January 1 to December 31, 2021 in the Tangshan People's Hospital (TS), Chongqing, China, and 245 lesions from 235 patients between January 1 to December 31, 2021 in the Dazu People's Hospital (DZ), Chongqing, China. All lesions in the training and total validation cohort were US benign findings during screening and biopsy-confirmed malignant, benign, and benign with 3-year follow-up records. Six radiologists performed the clinical diagnostic performance of EDL-BC, and six radiologists independently reviewed the retrospective datasets on a web-based rating platform. The area under the receiver operating characteristic curve (AUC) of the internal validation cohort and two independent external validation cohorts for EDL-BC was 0.950 (95% confidence interval [CI]: 0.909-0.969), 0.956 (95% [CI]: 0.939-0.971), and 0.907 (95% [CI]: 0.877-0.938), respectively. The sensitivity values were 94.4% (95% [CI]: 72.7%-99.9%), 100% (95% [CI]: 69.2%-100%), and 80% (95% [CI]: 28.4%-99.5%), respectively, at 0.76. The AUC for accurate diagnosis of EDL-BC (0.945 [95% [CI]: 0.933-0.965]) and radiologists with artificial intelligence (AI) assistance (0.899 [95% [CI]: 0.883-0.913]) was significantly higher than that of the radiologists without AI assistance (0.716 [95% [CI]: 0.693-0.738]; p<0.0001). Furthermore, there were no significant differences between the EDL-BC model and radiologists with AI assistance (p=0.099). EDL-BC can identify subtle but informative elements on US images of breast lesions and can significantly improve radiologists' diagnostic performance for identifying patients with early breast cancer and benefiting the clinical practice. The National Key R&D Program of China.

Machine Learning for Prediction of Adverse Cardiovascular Events in Adults With Repaired Tetralogy of Fallot Using Clinical and Cardiovascular Magnetic Resonance Imaging Variables

Background: Existing models for prediction of major adverse cardiovascular events (MACE) after repair of tetralogy of Fallot have been limited by modest predictive capacity and limited applicability to routine clinical practice. We hypothesized that an artificial intelligence model using an array of parameters would enhance 5-year MACE prediction in adults with repaired tetralogy of Fallot. Methods: A machine learning algorithm was applied to 2 nonoverlapping, institutional databases of adults with repaired tetralogy of Fallot: (1) for model development, a prospectively constructed clinical and cardiovascular magnetic resonance registry; (2) for model validation, a retrospective database comprised of variables extracted from the electronic health record. The MACE composite outcome included mortality, resuscitated sudden death, sustained ventricular tachycardia and heart failure. Analysis was restricted to individuals with MACE or followed ≥5 years. A random forest model was trained using machine learning (n=57 variables). Repeated random sub-sampling validation was sequentially applied to the development dataset followed by application to the validation dataset. Results: We identified 804 individuals (n=312 for development and n=492 for validation). Model prediction (area under the curve [95% CI]) for MACE in the validation dataset was strong (0.82 [0.74–0.89]) with superior performance to a conventional Cox multivariable model (0.63 [0.51–0.75]; P =0.003). Model performance did not change significantly with input restricted to the 10 strongest features (decreasing order of strength: right ventricular end-systolic volume indexed, right ventricular ejection fraction, age at cardiovascular magnetic resonance imaging, age at repair, absolute ventilatory anaerobic threshold, right ventricular end-diastolic volume indexed, ventilatory anaerobic threshold % predicted, peak aerobic capacity, left ventricular ejection fraction, and pulmonary regurgitation fraction; 0.81 [0.72–0.89]; P =0.232). Removing exercise parameters resulted in inferior model performance (0.75 [0.65–0.84]; P =0.002). Conclusions: In this single-center study, a machine learning-based prediction model comprised of readily available clinical and cardiovascular magnetic resonance imaging variables performed well in an independent validation cohort. Further study will determine the value of this model for risk stratification in adults with repared tetralogy of Fallot.

Early detection of lung cancer using a panel of circulating cell-free DNA methylation biomarkers.

e20614 Background: Lung cancer remains the leading cause of cancer mortality worldwide. Early detection of lung cancer helps improve treatment and survival. Numerous aberrant DNA methylations have been reported in early-stage lung cancer. Here, we sought to identify novel DNA methylation biomarkers that could potentially be used for noninvasive early diagnosis of lung cancers. Methods: This prospective-specimen collection and retrospective-blinded-evaluation trial enrolled a total of 317 participants (198 tissues and 119 plasmas) comprising healthy controls, patients with lung cancer and benign disease from The First Affiliated Hospital of Soochow University between January 2020 and December 2021. Tissue and plasma samples were subjected to targeted bisulfite sequencing with a lung cancer specific panel targeting 9,307 differential methylation regions (DMRs). DMRs associated with lung cancer were identified by comparing the methylation profiles of tissue samples from patients with lung cancer and benign disease. Markers were selected with minimum redundancy and maximum relevance (mRMR) algorithm. A prediction model for lung cancer diagnosis was built through logistic regression algorithm and validated independently in tissue samples. Furthermore, the performance of this developed model was evaluated in a set of plasma cell-free DNA (cfDNA) samples. Results: We identified 7 DMRs corresponding to 7 differentially methylated genes (DMRs, including HOXB4, HOXA7, HOXD8, ITGA4, ZNF808, PTGER4, and B3GNTL1), highly associated with lung cancer by comparing the methylation profiles of lung cancer and benign nodule tissues. Based on the 7-DMR biomarker panel, we developed a new diagnostic model in tissue samples, termed “7-DMR model”, to distinguish lung cancers from benign diseases, achieving AUCs of 0.966 (95%CI: 0.933-1.000)/0.961 (0.924-1.000), sensitivities of 0.887 (0.824-0.951)/0.922 (0.863-0.980), specificities of 0.938 (0.889-0.986)/1.000 (1.000-1.000), and accuracies of 0.896 (0.835-0.957)/0.938 (0.886-0.991) in the discovery cohort (n = 96) and the independent validation cohort (n = 81), respectively. Furthermore, the 7-DMR model was applied to noninvasive discrimination of lung cancers and non-lung cancers including benign lung diseases and healthy controls in an independent validation cohort of plasma samples (n = 106), yielding an AUC of 0.935 (0.862-1.000), sensitivity of 0.808 (0.733-0.883), specificity of 0.975 (0.945-1.000), and accuracy of 0.934 (0.887-0.981). Conclusions: The 7 novel DMRs could be promising methylation biomarkers that merits further development as a noninvasive test for early detection of lung cancer.

Tumor-intrinsic features associated with progression-free survival (PFS) in patients (pts) with relapsed and refractory multiple myeloma (RRMM) treated with idecabtagene vicleucel (ide-cel).

8035 Background: Ide-cel, a B-cell maturation antigen chimeric antigen receptor T cell therapy, has demonstrated frequent, deep, and durable responses in pts with RRMM in the KarMMa (NCT03361748), KarMMa-2 (NCT03601078), and KarMMa-3 (NCT03651128) trials. A previous analysis of KarMMa showed comparable efficacy of ide-cel across molecular high-risk/resistance (HR/R) features such as biallelic p53 inactivation, 1q amplification, t(4;14), and CRBN dysregulation, and identified a baseline gene expression pattern (PC4) associated with PFS (Martin N, et al. Hemasphere 2022;6:(S6):1452). We aim to identify additional genomic features associated with ide-cel efficacy in KarMMa and KarMMa-2, and evaluate samples from KarMMa-3 pts as an independent validation cohort. Methods: Transcriptional and genomic profiles were assessed at baseline in CD138+ cells from bone marrow from pts in KarMMa, KarMMa-2 cohort 1, and KarMMa-3. Gene copy number aberrations were evaluated for associations with response. Analyses were post hoc and exploratory, and a P-value, or false discovery rate (FDR), of &lt; 0.05 was used to identify associations of interest. Results: In evaluable pts in KarMMa (n = 70), single copy number loss was observed at 2 loci that associated with PFS, a broad region across 14q (n = 16) and deletion at 1p31.2 (n = 6). These were generally independent and collectively represented 30% (21/70) of pts. Both were associated with shorter median PFS (mPFS) in KarMMa (cohort mPFS, 8.8 mo, n = 128; Munshi et al. N Engl J Med 2021;384:705-16) after FDR correction (del 14q, mPFS, 4.0 mo; del 1p31.2, mPFS, 2.3 mo). No pt with 1p loss achieved a best overall response (BOR) better than partial response. Pts with loss vs non-loss of 14q had similar BOR distribution, but poorer mPFS within each BOR (10.4 vs 29.7 mo for loss vs non-loss in pts with complete response/stringent complete response). Due to small sample size in KarMMa-2 cohort 1 (n = 18), these data were aggregated with those from KarMMa. When combined, previously described associations with HR/R features were maintained except for biallelic p53 inactivation, which showed a stronger association with shorter PFS in the aggregate dataset (n = 83, P = 0.04) than in KarMMa (n = 65, P = 0.09). KarMMa-3 (n = 210) analysis is ongoing and will be used as a validation cohort for these findings. Consistent trends for the association of the PC4 gene signature with PFS were observed in KarMMa-2 but were not statistically significant and limited by sample size. Conclusions: Two novel genomic HR/R features associated with PFS were identified using samples from KarMMa; validation is ongoing using samples from KarMMa-3. The PC4 gene signature was consistent in 2 RRMM cohorts. A modest association between biallelic p53 disruption and PFS emerged with increased sample size and will be studied in KarMMa-3.