- Email: [email protected]
Enhanced performance of prognostic estimation from TCGA RNAseq data using transfer learning
abstracts 165P
Enhanced performance of prognostic estimation from TCGA RNAseq data using transfer learning
Background: Deep learning (DL) is one of the best approaches to predict nonlinear behaviors from high dimensional data. Nevertheless predicting the outcome of patients affected by cancers from transcriptomic data has shown limited performance, even with DL (C-index usually <0.65). Transfer learning is a DL two-step method where a model is pre-trained for a basic task on large amount of data, and then fine-tuned on the aimed task. We hypothesized that using TL with RNAseq may improve the performances of cancer patients’ outcome estimation. Methods: The model was a Multi-Mayer Perceptron (MLP) with 22913 inputs corresponding to genes bulk tumor whole genome RNAseq expression analysis. An important restriction was applied to the number of units at second layer (N ¼ 100), with further linear decrease across subsequent layers. Architecture of the model (number of layers, skip connections), L1 normalization value and learning rate were optimized by grid search on 30 parallel models. Training was performed using Keras package in R. Data were split into 70% training, 15% cross validation, 15% validation for each step, without contamination between the 2 transfer learning steps. The pre-training step consisted in predicting the organs of sample origin using 17.487 public RNAseq data of normal & cancer tissues (GTEX from gtexportal.org & TCGA from cBioportal.org). Fine-tuning on patients survival used 6401 training tumors. The model’s performance on survival prediction was evaluated by C-index and the area under the survival receiver-operating characteristic curve (AUROC). Results: The pre-training using GTEx and TCGA reached very high performance with validation accuracy of 0.96 to predict organ of origins for the best model (all models had validation accuracy > 0.9). Fine-tuning on survival, the prognostic performance of the best model on the validation cohort was C-index¼0.74 and AUROC¼ 0.81 (80% of models had a C-index > 0.6). The best model had 8 hidden layers and a small penalization value. Conclusions: Thanks to this original transfer learning method, we achieved a high performance to estimate cancer patients’ prognostic from whole genome expression, a classically challenging task. Learning on public databases is a valuable method of DL for personalized cancer care. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: E. Angevin: Advisory / Consultancy: Amgen; Advisory / Consultancy: Astellas; Advisory / Consultancy: AstraZeneca; Advisory / Consultancy: Bayer; Advisory / Consultancy: BeiGene; Advisory / Consultancy: BMS; Advisory / Consultancy: Celgene; Advisory / Consultancy: DebioPharma; Advisory / Consultancy: Genentech; Advisory / Consultancy: Ipsen; Advisory / Consultancy: Janssen; Advisory / Consultancy: Lilly; Advisory / Consultancy: MedImmune; Advisory / Consultancy: Novartis; Advisory / Consultancy: Pfizer; Advisory / Consultancy: Roche; Advisory / Consultancy: Sanofi; Advisory / Consultancy: Orion. A. Hollebecque: Advisory / Consultancy: Amgen; Advisory / Consultancy: Spectrum Pharmaceuticals; Advisory / Consultancy: Lilly; Advisory / Consultancy: Debiopharm; Travel / Accommodation / Expenses: Servier; Travel / Accommodation / Expenses: Amgen; Travel / Accommodation / Expenses: Lilly; Travel / Accommodation / Expenses: Incyte; Travel / Accommodation / Expenses: Debiopharm. E. Deutsch: Advisory / Consultancy: Boehringer; Advisory / Consultancy: Medimune; Advisory / Consultancy: Amgen; Research grant / Funding (self): AstraZeneca; Research grant / Funding (self): biotrachea; Research grant / Funding (institution): BristolMyersSquidd; Research grant / Funding (self): Clevelex; Research grant / Funding (self): EDF; Research grant / Funding (self): Lilly; Research grant / Funding (self): GlaxoSmisthKline; Research grant / Funding (self): Merk; Research grant / Funding (self): Nanobiotix; Research grant / Funding (self): Oseo; Research grant / Funding (self): Ray Search Laboratory; Research grant / Funding (self): Roche; Research grant / Funding (self): Ipsen; Research grant / Funding (self): Servier; Research grant / Funding (self): Takeda. C. Massard: Advisory / Consultancy: Amgen; Advisory / Consultancy: Astellas; Advisory / Consultancy: AstraZeneca; Advisory / Consultancy: Bayer; Advisory / Consultancy: BeiGene; Advisory / Consultancy: BMS; Advisory / Consultancy: Celgene; Advisory / Consultancy: DebioPharma; Advisory / Consultancy: Genentech; Advisory / Consultancy: Ipsen; Advisory / Consultancy: Janssen; Advisory / Consultancy: Lilly; Advisory / Consultancy: MedImmune; Advisory / Consultancy: Novartis; Advisory / Consultancy: Pfizer; Advisory / Consultancy: Roche; Advisory / Consultancy: Sanofi; Advisory / Consultancy: Orion. L. Verlingue: Research grant / Funding (self): Bristol-Myers Squibb; Advisory / Consultancy: Pierre Fabre; Advisory / Consultancy: Adaptherapy. All other authors have declared no conflicts of interest.
166P
Analysis of circulating tumour DNA for early relapse detection in stage III colorectal cancer after adjuvant chemotherapy
S.A. Jacobs1, H. Sethi2, T. Kolveska3, T.J. George1,4, S. Shchegrova2, T. Tin5, J. Lee5, A. Olson2, D. Renner2, E. Kalashnikova2, G. Yothers5,6, N. Wolmark1,5, K.L. Pogue-Geile1, A. Srinivasan1, J. Kortmansky7, M. Louie2, R. Salari2, B. Zimmermann2, A. Aleshin2, C.J. Allegra1,4 1 NSABP Foundation, Pittsburgh, PA, USA, 2Natera, Inc. San Carlos, CA, USA, 3Kaiser Permanente Oncology Clinical Trials Program, Vallejo, CA, USA, 4University of Florida Health Cancer Center, Gainesville, FL, USA, 5University of Pittsburgh, Pittsburgh, PA, USA, 6NRG Oncology, Pittsburgh, PA, USA, 7Yale, New Haven, CT, USA Background: Circulating tumor DNA (ctDNA) has emerged as a promising biomarker for early prediction of relapse across different tumor types. In patients with colorectal cancer (CRC), multiple studies have analyzed ctDNA to monitor tumor burden using fixed gene panels and droplet digital PCR. Here, we use a highly sensitive and specific, bespoke, whole exome-based NGS approach (SignateraTM) for ctDNA monitoring. Methods: A cohort of 33 patients with stage III CRC who underwent surgery and were treated with at least 4 months of adjuvant chemotherapy was analyzed. Mutational profiles derived from primary tumor tissue and germline DNA whole exome were used to design assays targeting tumor-specific somatic variants. The bespoke assays were used for ctDNA detection in plasma samples. Relapse-free survival (RFS) was calculated for patients stratified by ctDNA status. Results: Plasma samples (n ¼ 44; average volume¼1.8mL) from patients (N ¼ 33) were analysed for the presence of ctDNA. Of the five ctDNA-positive patients, clinical follow-up was available for three patients, all of whom relapsed (100%; 3/3); three of 27 ctDNA-negative patients (11%) also clinically relapsed. Molecular relapse through ctDNA analysis was detected up to 668 days ahead of radiological imaging with an average lead time of 305 days. The majority of relapses in ctDNA-positive patients (67%; 2/ 3) occurred within a year of follow-up, while no relapses were observed in ctDNA-negative patients during the one-year time frame. All plasma samples (n ¼ 34) from 24 nonrelapsing patients were ctDNA negative, corresponding to a specificity of 100%. The presence of ctDNA was associated with a markedly reduced RFS compared to ctDNA-negative patients (HR: 5.6; 95% CI: 0.6-52.1; p < 0.01). Conclusions: The study results indicate that ctDNA status is associated with high relapse risk in patients with CRC and can serve as a predictor of patient outcome. Despite low plasma volumes (<5mL) and lack of longitudinal samples for analysis, ctDNA was detected in 50% of relapse cases. Legal entity responsible for the study: Natera, Inc. and NSABP Foundation. Funding: Natera, Inc., Bayer, NSABP Foundation. Disclosure: H. Sethi: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. T.J. George: Research grant / Funding (institution): Merck; Research grant / Funding (institution): AstraZeneca; Research grant / Funding (institution): BMS; Research grant / Funding (institution): Bayer; Research grant / Funding (institution): Lilly; Research grant / Funding (institution): Incyte; Research grant / Funding (institution): Seattle Genetics; Research grant / Funding (institution): Pharmacyclics. S. Shchegrova: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A.S. Tin: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A. Olson: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. D. Renner: Full / Part-time employment: Natera, Inc. E. Kalashnikova: Full / Part-time employment: Natera, Inc. M. Louie: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. R. Salari: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. B. Zimmermann: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A. Aleshin: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. All other authors have declared no conflicts of interest.
167P
Ascites-derived circulating microRNAs as potential diagnostic biomarkers of gastric cancer-associated malignant ascites
H.S. Han1, H.B. Chae1, J. Yun2, H.J. Kim3, S-I. Go3, W.S. Lee3, W.K. Bae4, S.H. Cho4, E.-K. Song5 1 Department of of Internal Medicine, Chungbuk National University College of Medicine, Cheongju, Republic of Korea, 2Department of Pharmaceutical Engineering, College of Science Engineering, Cheongju University, Cheongju, Republic of Korea, 3 Department of Internal Medicine, Gyeongsang National University School of Medicine, Jinju, Republic of Korea, 4Department of Internal Medicine, Chonnam National University Medical School, Gwangju, Republic of Korea, 5Department of Internal Medicine, Chonbuk National University Medical School, Jeonju, Republic of Korea Background: Peritoneal carcinomatosis with malignant ascites is associated with dismal prognosis in gastric cancer. Malignant ascites is the most relevant body fluid in which to seek diagnostic biomarkers for peritoneal carcinomatosis. We aimed to identify and validate ascites-derived circulating microRNAs (miRNAs) that are differentially expressed between liver cirrhosis-associated benign ascites (LC-ascites) and gastric cancer-associated malignant ascites (GC-ascites). Methods: MiRNA expression levels were investigated in three independent cohorts. Overall, 165 ascites samples (73 LC-ascites and 92 GC-ascites) were obtained from the National Biobank of Korea. Initially, microarrays were used to screen the expression levels of 2,006 miRNAs in the discovery cohort (n ¼ 22). Subsequently, quantitative
v52 | Biomarkers
Volume 30 | Supplement 5 | October 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_5/mdz239.073/5577784 by Miami University user on 21 October 2019
H. Vanacker1, E. Angevin1, A. Hollebecque1, R. Sun2, E. Deutsch3, A. Zynovyev4, L. Calzone4, E. Barillot4, C. Massard1, L. Verlingue1 1 Drug Development Department (DITEP), Gustave Roussy - Cancer Campus, Villejuif, France, 2Department of Radiation Oncology, Gustave Roussy Cancer Campus, Villejuif, France, 3Department of Radiation Oncology, Radiomics Team, Molecular Radiotherapy INSERM U1030, University Paris-Saclay, Faculty of Medicine, Gustave Roussy Cancer Campus, Villejuif, France, 4U900 bioinformatics biostatistics and epidemiology, Institut Curie, Paris, France
Annals of Oncology
Enhanced performance of prognostic estimation from TCGA RNAseq data using transfer learning
Background: Deep learning (DL) is one of the best approaches to predict nonlinear behaviors from high dimensional data. Nevertheless predicting the outcome of patients affected by cancers from transcriptomic data has shown limited performance, even with DL (C-index usually <0.65). Transfer learning is a DL two-step method where a model is pre-trained for a basic task on large amount of data, and then fine-tuned on the aimed task. We hypothesized that using TL with RNAseq may improve the performances of cancer patients’ outcome estimation. Methods: The model was a Multi-Mayer Perceptron (MLP) with 22913 inputs corresponding to genes bulk tumor whole genome RNAseq expression analysis. An important restriction was applied to the number of units at second layer (N ¼ 100), with further linear decrease across subsequent layers. Architecture of the model (number of layers, skip connections), L1 normalization value and learning rate were optimized by grid search on 30 parallel models. Training was performed using Keras package in R. Data were split into 70% training, 15% cross validation, 15% validation for each step, without contamination between the 2 transfer learning steps. The pre-training step consisted in predicting the organs of sample origin using 17.487 public RNAseq data of normal & cancer tissues (GTEX from gtexportal.org & TCGA from cBioportal.org). Fine-tuning on patients survival used 6401 training tumors. The model’s performance on survival prediction was evaluated by C-index and the area under the survival receiver-operating characteristic curve (AUROC). Results: The pre-training using GTEx and TCGA reached very high performance with validation accuracy of 0.96 to predict organ of origins for the best model (all models had validation accuracy > 0.9). Fine-tuning on survival, the prognostic performance of the best model on the validation cohort was C-index¼0.74 and AUROC¼ 0.81 (80% of models had a C-index > 0.6). The best model had 8 hidden layers and a small penalization value. Conclusions: Thanks to this original transfer learning method, we achieved a high performance to estimate cancer patients’ prognostic from whole genome expression, a classically challenging task. Learning on public databases is a valuable method of DL for personalized cancer care. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: E. Angevin: Advisory / Consultancy: Amgen; Advisory / Consultancy: Astellas; Advisory / Consultancy: AstraZeneca; Advisory / Consultancy: Bayer; Advisory / Consultancy: BeiGene; Advisory / Consultancy: BMS; Advisory / Consultancy: Celgene; Advisory / Consultancy: DebioPharma; Advisory / Consultancy: Genentech; Advisory / Consultancy: Ipsen; Advisory / Consultancy: Janssen; Advisory / Consultancy: Lilly; Advisory / Consultancy: MedImmune; Advisory / Consultancy: Novartis; Advisory / Consultancy: Pfizer; Advisory / Consultancy: Roche; Advisory / Consultancy: Sanofi; Advisory / Consultancy: Orion. A. Hollebecque: Advisory / Consultancy: Amgen; Advisory / Consultancy: Spectrum Pharmaceuticals; Advisory / Consultancy: Lilly; Advisory / Consultancy: Debiopharm; Travel / Accommodation / Expenses: Servier; Travel / Accommodation / Expenses: Amgen; Travel / Accommodation / Expenses: Lilly; Travel / Accommodation / Expenses: Incyte; Travel / Accommodation / Expenses: Debiopharm. E. Deutsch: Advisory / Consultancy: Boehringer; Advisory / Consultancy: Medimune; Advisory / Consultancy: Amgen; Research grant / Funding (self): AstraZeneca; Research grant / Funding (self): biotrachea; Research grant / Funding (institution): BristolMyersSquidd; Research grant / Funding (self): Clevelex; Research grant / Funding (self): EDF; Research grant / Funding (self): Lilly; Research grant / Funding (self): GlaxoSmisthKline; Research grant / Funding (self): Merk; Research grant / Funding (self): Nanobiotix; Research grant / Funding (self): Oseo; Research grant / Funding (self): Ray Search Laboratory; Research grant / Funding (self): Roche; Research grant / Funding (self): Ipsen; Research grant / Funding (self): Servier; Research grant / Funding (self): Takeda. C. Massard: Advisory / Consultancy: Amgen; Advisory / Consultancy: Astellas; Advisory / Consultancy: AstraZeneca; Advisory / Consultancy: Bayer; Advisory / Consultancy: BeiGene; Advisory / Consultancy: BMS; Advisory / Consultancy: Celgene; Advisory / Consultancy: DebioPharma; Advisory / Consultancy: Genentech; Advisory / Consultancy: Ipsen; Advisory / Consultancy: Janssen; Advisory / Consultancy: Lilly; Advisory / Consultancy: MedImmune; Advisory / Consultancy: Novartis; Advisory / Consultancy: Pfizer; Advisory / Consultancy: Roche; Advisory / Consultancy: Sanofi; Advisory / Consultancy: Orion. L. Verlingue: Research grant / Funding (self): Bristol-Myers Squibb; Advisory / Consultancy: Pierre Fabre; Advisory / Consultancy: Adaptherapy. All other authors have declared no conflicts of interest.
166P
Analysis of circulating tumour DNA for early relapse detection in stage III colorectal cancer after adjuvant chemotherapy
S.A. Jacobs1, H. Sethi2, T. Kolveska3, T.J. George1,4, S. Shchegrova2, T. Tin5, J. Lee5, A. Olson2, D. Renner2, E. Kalashnikova2, G. Yothers5,6, N. Wolmark1,5, K.L. Pogue-Geile1, A. Srinivasan1, J. Kortmansky7, M. Louie2, R. Salari2, B. Zimmermann2, A. Aleshin2, C.J. Allegra1,4 1 NSABP Foundation, Pittsburgh, PA, USA, 2Natera, Inc. San Carlos, CA, USA, 3Kaiser Permanente Oncology Clinical Trials Program, Vallejo, CA, USA, 4University of Florida Health Cancer Center, Gainesville, FL, USA, 5University of Pittsburgh, Pittsburgh, PA, USA, 6NRG Oncology, Pittsburgh, PA, USA, 7Yale, New Haven, CT, USA Background: Circulating tumor DNA (ctDNA) has emerged as a promising biomarker for early prediction of relapse across different tumor types. In patients with colorectal cancer (CRC), multiple studies have analyzed ctDNA to monitor tumor burden using fixed gene panels and droplet digital PCR. Here, we use a highly sensitive and specific, bespoke, whole exome-based NGS approach (SignateraTM) for ctDNA monitoring. Methods: A cohort of 33 patients with stage III CRC who underwent surgery and were treated with at least 4 months of adjuvant chemotherapy was analyzed. Mutational profiles derived from primary tumor tissue and germline DNA whole exome were used to design assays targeting tumor-specific somatic variants. The bespoke assays were used for ctDNA detection in plasma samples. Relapse-free survival (RFS) was calculated for patients stratified by ctDNA status. Results: Plasma samples (n ¼ 44; average volume¼1.8mL) from patients (N ¼ 33) were analysed for the presence of ctDNA. Of the five ctDNA-positive patients, clinical follow-up was available for three patients, all of whom relapsed (100%; 3/3); three of 27 ctDNA-negative patients (11%) also clinically relapsed. Molecular relapse through ctDNA analysis was detected up to 668 days ahead of radiological imaging with an average lead time of 305 days. The majority of relapses in ctDNA-positive patients (67%; 2/ 3) occurred within a year of follow-up, while no relapses were observed in ctDNA-negative patients during the one-year time frame. All plasma samples (n ¼ 34) from 24 nonrelapsing patients were ctDNA negative, corresponding to a specificity of 100%. The presence of ctDNA was associated with a markedly reduced RFS compared to ctDNA-negative patients (HR: 5.6; 95% CI: 0.6-52.1; p < 0.01). Conclusions: The study results indicate that ctDNA status is associated with high relapse risk in patients with CRC and can serve as a predictor of patient outcome. Despite low plasma volumes (<5mL) and lack of longitudinal samples for analysis, ctDNA was detected in 50% of relapse cases. Legal entity responsible for the study: Natera, Inc. and NSABP Foundation. Funding: Natera, Inc., Bayer, NSABP Foundation. Disclosure: H. Sethi: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. T.J. George: Research grant / Funding (institution): Merck; Research grant / Funding (institution): AstraZeneca; Research grant / Funding (institution): BMS; Research grant / Funding (institution): Bayer; Research grant / Funding (institution): Lilly; Research grant / Funding (institution): Incyte; Research grant / Funding (institution): Seattle Genetics; Research grant / Funding (institution): Pharmacyclics. S. Shchegrova: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A.S. Tin: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A. Olson: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. D. Renner: Full / Part-time employment: Natera, Inc. E. Kalashnikova: Full / Part-time employment: Natera, Inc. M. Louie: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. R. Salari: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. B. Zimmermann: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. A. Aleshin: Shareholder / Stockholder / Stock options, Full / Part-time employment, I am an employee of Natera and own stock/options to stock.: Natera, Inc. All other authors have declared no conflicts of interest.
167P
Ascites-derived circulating microRNAs as potential diagnostic biomarkers of gastric cancer-associated malignant ascites
H.S. Han1, H.B. Chae1, J. Yun2, H.J. Kim3, S-I. Go3, W.S. Lee3, W.K. Bae4, S.H. Cho4, E.-K. Song5 1 Department of of Internal Medicine, Chungbuk National University College of Medicine, Cheongju, Republic of Korea, 2Department of Pharmaceutical Engineering, College of Science Engineering, Cheongju University, Cheongju, Republic of Korea, 3 Department of Internal Medicine, Gyeongsang National University School of Medicine, Jinju, Republic of Korea, 4Department of Internal Medicine, Chonnam National University Medical School, Gwangju, Republic of Korea, 5Department of Internal Medicine, Chonbuk National University Medical School, Jeonju, Republic of Korea Background: Peritoneal carcinomatosis with malignant ascites is associated with dismal prognosis in gastric cancer. Malignant ascites is the most relevant body fluid in which to seek diagnostic biomarkers for peritoneal carcinomatosis. We aimed to identify and validate ascites-derived circulating microRNAs (miRNAs) that are differentially expressed between liver cirrhosis-associated benign ascites (LC-ascites) and gastric cancer-associated malignant ascites (GC-ascites). Methods: MiRNA expression levels were investigated in three independent cohorts. Overall, 165 ascites samples (73 LC-ascites and 92 GC-ascites) were obtained from the National Biobank of Korea. Initially, microarrays were used to screen the expression levels of 2,006 miRNAs in the discovery cohort (n ¼ 22). Subsequently, quantitative
v52 | Biomarkers
Volume 30 | Supplement 5 | October 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_5/mdz239.073/5577784 by Miami University user on 21 October 2019
H. Vanacker1, E. Angevin1, A. Hollebecque1, R. Sun2, E. Deutsch3, A. Zynovyev4, L. Calzone4, E. Barillot4, C. Massard1, L. Verlingue1 1 Drug Development Department (DITEP), Gustave Roussy - Cancer Campus, Villejuif, France, 2Department of Radiation Oncology, Gustave Roussy Cancer Campus, Villejuif, France, 3Department of Radiation Oncology, Radiomics Team, Molecular Radiotherapy INSERM U1030, University Paris-Saclay, Faculty of Medicine, Gustave Roussy Cancer Campus, Villejuif, France, 4U900 bioinformatics biostatistics and epidemiology, Institut Curie, Paris, France
Annals of Oncology