- Email: [email protected]
A Bayesian framework to quantify survival uncertainty
abstracts Conclusions: Our newly developed targeted proximity ligation assays enable robust detection of structural variation in FFPE samples and the development of personalized MRD/ctDNA tests. Legal entity responsible for the study: Cergentis. Funding: Cergentis via Horizon2020 SME funding. Disclosure: H. Feitsma: Full / Part-time employment: Cergentis. M. Yilmaz: Full / Part-time employment: Cergentis. J. Swennenhuis: Full / Part-time employment: Cergentis. A. Rakszewska: Full / Part-time employment: Cergentis. K. Hajo: Full / Part-time employment: Cergentis. E. Splinter: Full / Part-time employment: Cergentis. M. Simonis: Full / Part-time employment: Cergentis. M. Van Min: Shareholder / Stockholder / Stock options, Full / Part-time employment: Cergentis. All other authors have declared no conflicts of interest.
Association of rs363293 single nucleotide polymorphism in promoter region of miRNA-143/145 with susceptibility to colorectal cancer and with patients’ outcome
G.M. Mihaylova1, D. Ivanova1, R. Manev1, O. Tasinov1, N. Nazifova-Tasinova1, V. Petrova2, D. Petkova- Nelova2, Z. Mihaylova2, N. Conev1, I. Donev3, M. Radanova1 1 Medical University of Varna, Varna, Bulgaria, 2Military Medical Academy, Sofia, Bulgaria, 3Hospital Nadezhda, Sofia, Bulgaria Background: miRNA-143 and miRNA-145 are known as suppressors of tumor cell proliferation and migration, and in triggering of apoptosis. Both miRNAs have been shown to be down-regulated in colorectal cancer (CRC). miRNA-143 and miRNA-145 are encoded from gene cluster on chromosome 5q32 and probably originate from the same primary RNA. The aim of the present study was to determine the possible association of gene variant in promoter region of miRNA-143/145 with risk of CRC in a Bulgarian population and to evaluate the role of the rs363293 single nucleotide polymorphism (SNP) as potential prognostic and/or predictive biomarker of the disease. Methods: 99 patients with CRC and 89 healthy volunteers, all Caucasians from Bulgaria, were genotyped for rs363293 SNP by qPCR method. The expression levels of miRNA-143 and miRNA-145 in sera also have been determined. Relative gene expression was calculated using 2-DDCt method. Results: AA genotype in rs363293 SNP was significantly overrepresented in CRC patients when compared with healthy volunteers (32% vs 19%, OR ¼ 2.02, 95% CI: 1.03–3.98, p ¼ 0.041). Patients with AA genotype in rs363293 SNP had significantly (p ¼ 0.033) longer mean overall survival (OS) of 21.9 months (95% CI: 17.72–26.12) as compared to those with CT and CC genotypes – 16.6 months (95% CI: 13.72–19.51). In the multivariate analysis, AA genotype in rs363293 SNP showed a trend towards better outcome (HR ¼ 0.59, 95% CI: 0.33–1.08, p ¼ 0.086). The association results for genotypes in rs363293 SNP and reduced expression levels of miR-143 and/or miR-145 in sera were not significant. Conclusions: Our data suggest the implication of the A allele in rs363293 SNP as a risk factor for CRC in a homozygous status for Caucasians or at least for a Bulgarian population and for patients - AA genotype has prognostic significance. Legal entity responsible for the study: The authors. Funding: Medical University of Varna (Grant number FMS-53/18.12.2017) and Bulgarian National Science Fund (Grant number RG-06-H23/6, 18.12.2018). Disclosure: All authors have declared no conflicts of interest.
110P
TFs and positive correlation was revealed between mRNA levels of 3 TFs and 5 identified genes (rs > 0.5 for all the pairs in 3 cancer types). Conclusions: We revealed that IQGAP3, KIF4A, KIF18B, NCAPH, and TROAP genes are characterized by a drastic upregulation in 10 cancer types, including lung, breast, and colon cancers. These genes are potential “universal” oncogenes and possibly play a crucial role in carcinogenesis. The major mechanism of their upregulation is likely to be the activation by E2F1, FOXM1, and MYBL transcription factors. The revealed genes are potential therapeutic targets in multiple cancer types. Legal entity responsible for the study: Alexey A. Dmitriev. Funding: Russian Science Foundation (Grant 17-74-20064). Disclosure: All authors have declared no conflicts of interest.
Pan-cancer analysis for oncogenes and mechanisms of their upregulation
A.A. Dmitriev, G.S. Krasnov, N.V. Melnikova, E.N. Pushkova, R.O. Novakovskiy, A.V. Kudryavtseva, A.D. Beniaminov Engelhardt Institute of Molecular Biology Russian Academy of Sciences, Moscow, Russian Federation Background: Each tumor has a unique molecular portrait, nevertheless, there are also a number of genetic and epigenetic aberrations that are inherent to the groups of epithelial tumors of different localizations. Our research was aimed at the identification of genes with significantly increased expression (potential oncogenes) in multiple cancer types and determination of mechanisms of their upregulation. Methods: Bioinformatics analysis was performed using our CrossHub tool and The Cancer Genome Atlas (TCGA) transcriptomic data for 10 cancer types with at least 20 paired (tumor/normal) samples – breast, lung (AD and SCC), kidney (ccRCC and pRCC), thyroid, prostate, stomach, liver, and colon cancers. For qPCR analysis of gene expression, we used paired samples of non-small cell lung, breast, and colon cancers (40 for each). Results: Using bioinformatics analysis, we identified a number of genes that were significantly upregulated in 10 cancer types and chose 5 poorly studied for further research: IQGAP3, KIF4A, KIF18B, NCAPH, and TROAP. The qPCR analysis in a cohort of Russian patients showed expression increase of all 5 genes in at least 80% of samples with a 4-fold median increase in colon cancer and more than a 10-fold median increase in breast and lung cancers. Next, with the CrossHub tool and TCGA data, we predicted that the most likely mechanism of upregulation of identified genes in lung, breast, and colon cancers is activation of E2F1, FOXM1, and MYBL transcription factors (TFs). Using qPCR, strong expression increase was shown for genes encoding these
vii32 | Molecular Analysis for Personalised Therapy
111P
mTOR inhibitors in combination regimens guided by encyclopedic tumour analysis show superior outcomes compared to monotherapy in refractory cancers
T. Crook1, A. Vaid2, S. Limaye3, R. Page4, D. Patil5, D. Akolkar5, V. Datta5, A. Ghaisas5, R. Patil5, H. Singh5, A. Srinivasan5, S. Apurwa5, R. Datar5 1 Royal Surrey County Hospital, Guildford, UK, 2Medical and Haemato Oncology, Medanta - The Medicity, Gurugram, India, 3Medical Oncology, Kokilaben Dhirubhai Ambani Hospital, Mumbai, India, 4Biomedical Engineering, Worcester Polytechnic Institute, Worcester, MA, USA, 5Research and Innovations Department, Datar Cancer Genetics, Nashik, India Background: Though mTOR inhibition is considered an attractive strategy for cancer management, anti-mTOR monotherapies have not shown meaningful benefits. We hypothesized that an Encyclopedic Tumor Analysis (ETA) can identify vulnerabilities in the tumor in addition to mTOR activation. We further hypothesized that tandem synergistic targeting of these vulnerabilities using combination of mTOR inhibitors and other systemic anticancer agents in a label- and organ-agnostic manner can improve outcomes in refractory solid organ cancers as compared to mTOR inhibition monotherapy alone. Methods: Molecular Profiling (MP) of patients’ fresh tumor tissue interrogated gene alterations and differentially regulated metabolic pathways to identify druggable molecular targets in a label-agnostic manner. Immunohistochemistry (IHC) identified targetable hormone receptors (HR). Chemoresistance and response (CRR) profiling of viable tumor derived cells (TDCs) identified vulnerabilities of the tumor against a panel of systemic anticancer agents. Molecular indications linked to PIK3CA, mTOR, PTEN or TP53 genes were used for selection of mTOR inhibitors. Synergistic integration of MP, IHC and CRR datasets (i.e., ETA) generated patient-specific drug priority lists with projected efficacy and safety. Patients who received such ETA-guided treatments were evaluated by PET-CT scan to determine treatment response. Results: Among 41 patients who received combination treatments, 23 patients showed PR (ORR ¼ 56.1%), 16 showed SD (DCR ¼ 95.1%) and progression was observed in 2 patients. One patient who received monotherapy progressed at 27 days. Median PFS was 110 days (range 27 to 592). In the SHIVA trial where patients with mTOR activation (n ¼ 46) received monotherapy with mTOR inhibitor, median PFS of 72 days (range 57 to 100) was reported. No significant therapy–related adverse events were reported in any patient. Most patients reported stable to improved Quality of Life (QoL). Conclusions: ETA-guided combination regimens with mTOR inhibitors offer a viable and efficient strategy in advanced refractory malignancies and outperform mTOR inhibitor monotherapy. Legal entity responsible for the study: The authors. Funding: Datar Cancer Genetics Limited. Disclosure: D. Patil: Full / Part-time employment: Datar Cancer Genetics Limited. D. Akolkar: Full / Part-time employment: Datar Cancer Genetics Limited. V. Datta: Full / Part-time employment: Datar Cancer Genetics Limited. A. Ghaisas: Full / Part-time employment: Datar Cancer Genetics Limited. R. Patil: Full / Part-time employment: Datar Cancer Genetics Limited. H. Singh: Full / Parttime employment: Datar Cancer Genetics Limited. A. Srinivasan: Full / Part-time employment: Datar Cancer Genetics Limited. S. Apurwa: Full / Part-time employment: Datar Cancer Genetics Limited. R. Datar: Leadership role, Shareholder / Stockholder / Stock options, Officer / Board of Directors: Datar Cancer Genetics Limited. All other authors have declared no conflicts of interest.
112P
A Bayesian framework to quantify survival uncertainty
H. Loya1, D. Anand1, P. Poduval1, N. Kumar2, A. Sethi2 Department of Electrical Engineering, Indian Institute of Technology Bombay, Mumbai, India, 2Department of Pathology, University of Illinois at Chicago, Chicago, IL, USA
1
Background: Existing survival models do not model uncertainty in personalized (perpatient) prediction of survival, which is useful for treatment planning of cancers. Moreover, their restrictive modeling assumptions limit their accuracy of personalized survival estimation. For example, the Cox proportional hazard (CPH) model assumes a constant effect of each input feature on survival rate over time along with a linear combination of features. The multi-task logistic regression (MTLR) allows a smooth timevarying impact of each feature without relaxing the linearity assumption. Recently,
Volume 30 | Supplement 7 | November 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_7/mdz413.116/5613253 by guest on 30 December 2019
109P
Annals of Oncology
abstracts
Annals of Oncology
Table: 112P Comparison of models Methods
C-index mean (std. dev.)
IBS mean (std. dev.)
CPH MTLR N-MTLR B-MTLR BN-MTLR
0.67 (0.10) 0.68 (0.06) 0.68 (0.02) 0.69 (0.05) 0.70 (0.05)
0.20 (0.07) 0.21 (0.06) 0.16 (0.04) 0.14 (0.02) 0.10 (0.01)
Conclusions: The proposed Bayesian extensions of survival models give better mean personalized accuracy and allow computation of personalized uncertainty scores, which will pave the way for more informative models for treatment planning of cancers. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: All authors have declared no conflicts of interest.
114P
S. Shooter, J. Czarnecki, S. Nik-Zainal Medical Genetics, University of Cambridge - MRC Cancer Unit, Cambridge, UK Background: The somatic mutations found through sequencing a whole tumour genome are the aggregate outcome of one or multiple mutational processes. Each of these processes leaves a characteristic imprint, or mutational signature, on the genome. The final mutational profile is determined by the pattern of exposure of each of these mutational processes. Deconstructing the signature composition of a tumour’s mutational profile provides insights into the biological processes that drive it. The initial research in this field extracted 5 substitution signatures from 21 breast cancer samples, but research since has extracted signatures from ever expanding datasets of whole tumour genomes and experimental mutagenesis systems. Signatures have also been developed for different mutation types. The field has great potential to aid in the diagnosis and treatment of cancer patients. Methods: We developed a website using the ReactJS framework supported by a distributed compute cluster in OpenStack. Results: Here we present Signal, a new website enabling exploration of the latest research in the field of mutational signatures. With Signal we intend to inform the scientific community of the power of mutational signatures and to provide the results of our research in a dynamic, easy-to-use interface. This includes tissue-specific substitution and rearrangement signatures extracted from the PanCan dataset, in addition to signatures extracted from cell-lines exposed to environmental mutagens and geneknockouts. Users can upload their own variant calls for analysis by our pipeline, which will filter out kataegis regions, construct a substitution profile and detect the presence of any of our gold-standard substitution signatures. This signature fitting algorithm leverages the organ-specificity of our new signature extraction framework, leading to more accurate signature predictions than previously possible. Conclusions: We intend for Signal to be the home page of mutational signatures which will be kept at the forefront of research with the most recent published data and analysis methodologies. Legal entity responsible for the study: The authors. Funding: CRUK. Disclosure: All authors have declared no conflicts of interest.
115P 113P
Copy number variations in critical cell-signalling genes with potential targeted therapeutic application
E. Nikolova, V. Mitev, A. Todorova Medical Chemistry and Biochemistry, Medical University Sofia, Sofia, Bulgaria Background: Brain tumours are among “hard-to treat” world malignancy with the median survival time of 15 months for the most aggressive form - Glioblastoma multiforme (GBM). It is extremely complex and heterogeneous brain cancer, characterized by rapid progression, infiltration and therapeutic resistance. Currently, many trials for target therapy are focused on understanding the molecular heterogeneity of GBM and designing molecules that inhibit most common genetic alterations. Methods: In the present study, somatic copy number variations (CNVs) in critical signalling genes suitable for future target therapy are examined among Bulgarian patients. Fresh brain biopsies were collected from patients with primary and secondary GBM. Genomic DNA was isolated and CNVs were analysed by Multiplex ligation – dependent probe amplification (MLPA). Results: CDKN2A deletion is the most frequent genetic alteration (78%) among Bulgarian patients with GBM. The loss of this gene is a progression-associated genetic marker. Nearly 60% of GBM patients showed EGFR amplification. It is a prognostic and predictive marker and some patients with this aberration are suggested to respond better to EGFR inhibitors in combination therapy. PTEN deletion is another common event with the incidence rate of 33% and its loss corelates with the choice of mTOR and MEK inhibitors, as well as, the resistance to some EGFR inhibitors. TP53 deletion was found in only 11% of the analysed patients which is a positive marker and the presence of wild-type TP53 is targeted by a MDM2 inhibitors. In addition, MDM2 amplification was observed in one patient and this event also serve as a potential target by MDM2 inhibitors.Amplification in PDGFRA is observed in one patient. Depending on the status of other receptor tyrosine kinases, it could be targeted by PDGFR inhibitor and is also suggested to be a poor prognostic marker. Conclusions: Personalized treatment, customized for an individual patient genetics, has the potential to improve the therapy response. CNVs data could be useful for designing more effective personalized therapy. This is a small study on CNVs with potentially targeted therapeutic application in GBM for Bulgaria and recruitment of more patients is in progress. Legal entity responsible for the study: The authors. Funding: Medical University Sofia, Grant number D-127/2019. Disclosure: All authors have declared no conflicts of interest.
Volume 30 | Supplement 7 | November 2019
Signal: The home page of mutational signatures
Detection of actionable variants in various cancer types reveals value of whole-genome sequencing over in-silico whole-exome and hotspot panel sequencing
K.P. Ramarao-Milne, A-M. Patch, K. Nones, R. Koufariotis, F. Newell, V.R. Addala, O. Kondrashova, P. Mukhopadhyay, S.H. Kazakoff, V. Lakis, O. Holmes, C. Leonard, S. Wood, C. Xu, J.V. Pearson, G. Hollway, N. Waddell Medical Genomics, QIMR Berghofer Medical Research Institute, Brisbane, Australia Background: Declining costs of next-generation sequencing technologies has led to their more widespread use for the identification of targetable mutations. Nonetheless, while multigene panels are being assimilated into clinical practice, the use of wholeexome and whole-genome sequencing (WES and WGS) remains limited, as its added value is unclear in the clinic. Methods: In our study, we have used the Cancer Genome Interpreter (CGI) to identify the actionable variants that are detected by WGS and compared this to in-silico downsampled regions of WES and a hotspot panel. Results: We show that within some tumour types, WGS is invaluable in identifying actionable variants that involve copy number defects and structural aberrations. Moreover, WGS successfully identifies many resistance biomarkers that would have otherwise been missed by WES and panels, particularly in the breast cancer dataset studied. WGS is more valuable of the detection of actionable variants in some tumour types, whereas hotspot panels perform well for other tumour types. Additionally, we show that a tumour type-agnostic approach to selecting genomic biomarker-based drug allocation increases the number of possible FDA-approved and clinical trialbased drug prescriptions for any given genomic biomarker in the patient cohorts studied. Conclusions: Taken together, we show that WGS has the potential to impact patient care by identifying more targetable mutations and therefore expand drug options available for the patient. Furthermore, we illustrate the potential of tumour type-agnostic drug repurposing in nine cancer datasets. Legal entity responsible for the study: QIMR Berghofer Medical Research Institute. Funding: QIMR Berghofer Medical Research Institute. Disclosure: All authors have declared no conflicts of interest.
doi:10.1093/annonc/mdz413 | vii33
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_7/mdz413.116/5613253 by guest on 30 December 2019
neural-MTLR (N-MTLR) was proposed to relax the linearity assumption as well. We propose Bayesian (B-MTLR) and Bayesian neural network (BN-MTLR) frameworks for predicting survival that are more accurate than the previous methods and compute uncertainty for each personalized prediction. We demonstrate that including more granular information, such as PAM50 gene expressions instead of the PAM50 labels, both increases accuracy and decreases uncertainty in personalized survival prediction. Methods: Using four-fold cross validation on the TCGA-BRCA we compared the accuracy of personalized prediction (C-index and IBS) for CPH, MTLR, N-MTLR, BMTLR, and BN-MTLR using PAM50 labels and clinical variables. We also compared BN-MTLR models trained on PAM50 gene-set and clinical variables for decrease in uncertainty due to more granular inputs. Results: The table below shows that for PAM50 labels and clinical variables the C-index is higher and the IBS is lower for B-MTLR and BN-MTLR compared to the existing techniques. Additionally, BN-MTLR gave a mean uncertainty of 0.12 using PAM50 gene expression as opposed to 0.22 using PAM50 labels across all patients.
Association of rs363293 single nucleotide polymorphism in promoter region of miRNA-143/145 with susceptibility to colorectal cancer and with patients’ outcome
G.M. Mihaylova1, D. Ivanova1, R. Manev1, O. Tasinov1, N. Nazifova-Tasinova1, V. Petrova2, D. Petkova- Nelova2, Z. Mihaylova2, N. Conev1, I. Donev3, M. Radanova1 1 Medical University of Varna, Varna, Bulgaria, 2Military Medical Academy, Sofia, Bulgaria, 3Hospital Nadezhda, Sofia, Bulgaria Background: miRNA-143 and miRNA-145 are known as suppressors of tumor cell proliferation and migration, and in triggering of apoptosis. Both miRNAs have been shown to be down-regulated in colorectal cancer (CRC). miRNA-143 and miRNA-145 are encoded from gene cluster on chromosome 5q32 and probably originate from the same primary RNA. The aim of the present study was to determine the possible association of gene variant in promoter region of miRNA-143/145 with risk of CRC in a Bulgarian population and to evaluate the role of the rs363293 single nucleotide polymorphism (SNP) as potential prognostic and/or predictive biomarker of the disease. Methods: 99 patients with CRC and 89 healthy volunteers, all Caucasians from Bulgaria, were genotyped for rs363293 SNP by qPCR method. The expression levels of miRNA-143 and miRNA-145 in sera also have been determined. Relative gene expression was calculated using 2-DDCt method. Results: AA genotype in rs363293 SNP was significantly overrepresented in CRC patients when compared with healthy volunteers (32% vs 19%, OR ¼ 2.02, 95% CI: 1.03–3.98, p ¼ 0.041). Patients with AA genotype in rs363293 SNP had significantly (p ¼ 0.033) longer mean overall survival (OS) of 21.9 months (95% CI: 17.72–26.12) as compared to those with CT and CC genotypes – 16.6 months (95% CI: 13.72–19.51). In the multivariate analysis, AA genotype in rs363293 SNP showed a trend towards better outcome (HR ¼ 0.59, 95% CI: 0.33–1.08, p ¼ 0.086). The association results for genotypes in rs363293 SNP and reduced expression levels of miR-143 and/or miR-145 in sera were not significant. Conclusions: Our data suggest the implication of the A allele in rs363293 SNP as a risk factor for CRC in a homozygous status for Caucasians or at least for a Bulgarian population and for patients - AA genotype has prognostic significance. Legal entity responsible for the study: The authors. Funding: Medical University of Varna (Grant number FMS-53/18.12.2017) and Bulgarian National Science Fund (Grant number RG-06-H23/6, 18.12.2018). Disclosure: All authors have declared no conflicts of interest.
110P
TFs and positive correlation was revealed between mRNA levels of 3 TFs and 5 identified genes (rs > 0.5 for all the pairs in 3 cancer types). Conclusions: We revealed that IQGAP3, KIF4A, KIF18B, NCAPH, and TROAP genes are characterized by a drastic upregulation in 10 cancer types, including lung, breast, and colon cancers. These genes are potential “universal” oncogenes and possibly play a crucial role in carcinogenesis. The major mechanism of their upregulation is likely to be the activation by E2F1, FOXM1, and MYBL transcription factors. The revealed genes are potential therapeutic targets in multiple cancer types. Legal entity responsible for the study: Alexey A. Dmitriev. Funding: Russian Science Foundation (Grant 17-74-20064). Disclosure: All authors have declared no conflicts of interest.
Pan-cancer analysis for oncogenes and mechanisms of their upregulation
A.A. Dmitriev, G.S. Krasnov, N.V. Melnikova, E.N. Pushkova, R.O. Novakovskiy, A.V. Kudryavtseva, A.D. Beniaminov Engelhardt Institute of Molecular Biology Russian Academy of Sciences, Moscow, Russian Federation Background: Each tumor has a unique molecular portrait, nevertheless, there are also a number of genetic and epigenetic aberrations that are inherent to the groups of epithelial tumors of different localizations. Our research was aimed at the identification of genes with significantly increased expression (potential oncogenes) in multiple cancer types and determination of mechanisms of their upregulation. Methods: Bioinformatics analysis was performed using our CrossHub tool and The Cancer Genome Atlas (TCGA) transcriptomic data for 10 cancer types with at least 20 paired (tumor/normal) samples – breast, lung (AD and SCC), kidney (ccRCC and pRCC), thyroid, prostate, stomach, liver, and colon cancers. For qPCR analysis of gene expression, we used paired samples of non-small cell lung, breast, and colon cancers (40 for each). Results: Using bioinformatics analysis, we identified a number of genes that were significantly upregulated in 10 cancer types and chose 5 poorly studied for further research: IQGAP3, KIF4A, KIF18B, NCAPH, and TROAP. The qPCR analysis in a cohort of Russian patients showed expression increase of all 5 genes in at least 80% of samples with a 4-fold median increase in colon cancer and more than a 10-fold median increase in breast and lung cancers. Next, with the CrossHub tool and TCGA data, we predicted that the most likely mechanism of upregulation of identified genes in lung, breast, and colon cancers is activation of E2F1, FOXM1, and MYBL transcription factors (TFs). Using qPCR, strong expression increase was shown for genes encoding these
vii32 | Molecular Analysis for Personalised Therapy
111P
mTOR inhibitors in combination regimens guided by encyclopedic tumour analysis show superior outcomes compared to monotherapy in refractory cancers
T. Crook1, A. Vaid2, S. Limaye3, R. Page4, D. Patil5, D. Akolkar5, V. Datta5, A. Ghaisas5, R. Patil5, H. Singh5, A. Srinivasan5, S. Apurwa5, R. Datar5 1 Royal Surrey County Hospital, Guildford, UK, 2Medical and Haemato Oncology, Medanta - The Medicity, Gurugram, India, 3Medical Oncology, Kokilaben Dhirubhai Ambani Hospital, Mumbai, India, 4Biomedical Engineering, Worcester Polytechnic Institute, Worcester, MA, USA, 5Research and Innovations Department, Datar Cancer Genetics, Nashik, India Background: Though mTOR inhibition is considered an attractive strategy for cancer management, anti-mTOR monotherapies have not shown meaningful benefits. We hypothesized that an Encyclopedic Tumor Analysis (ETA) can identify vulnerabilities in the tumor in addition to mTOR activation. We further hypothesized that tandem synergistic targeting of these vulnerabilities using combination of mTOR inhibitors and other systemic anticancer agents in a label- and organ-agnostic manner can improve outcomes in refractory solid organ cancers as compared to mTOR inhibition monotherapy alone. Methods: Molecular Profiling (MP) of patients’ fresh tumor tissue interrogated gene alterations and differentially regulated metabolic pathways to identify druggable molecular targets in a label-agnostic manner. Immunohistochemistry (IHC) identified targetable hormone receptors (HR). Chemoresistance and response (CRR) profiling of viable tumor derived cells (TDCs) identified vulnerabilities of the tumor against a panel of systemic anticancer agents. Molecular indications linked to PIK3CA, mTOR, PTEN or TP53 genes were used for selection of mTOR inhibitors. Synergistic integration of MP, IHC and CRR datasets (i.e., ETA) generated patient-specific drug priority lists with projected efficacy and safety. Patients who received such ETA-guided treatments were evaluated by PET-CT scan to determine treatment response. Results: Among 41 patients who received combination treatments, 23 patients showed PR (ORR ¼ 56.1%), 16 showed SD (DCR ¼ 95.1%) and progression was observed in 2 patients. One patient who received monotherapy progressed at 27 days. Median PFS was 110 days (range 27 to 592). In the SHIVA trial where patients with mTOR activation (n ¼ 46) received monotherapy with mTOR inhibitor, median PFS of 72 days (range 57 to 100) was reported. No significant therapy–related adverse events were reported in any patient. Most patients reported stable to improved Quality of Life (QoL). Conclusions: ETA-guided combination regimens with mTOR inhibitors offer a viable and efficient strategy in advanced refractory malignancies and outperform mTOR inhibitor monotherapy. Legal entity responsible for the study: The authors. Funding: Datar Cancer Genetics Limited. Disclosure: D. Patil: Full / Part-time employment: Datar Cancer Genetics Limited. D. Akolkar: Full / Part-time employment: Datar Cancer Genetics Limited. V. Datta: Full / Part-time employment: Datar Cancer Genetics Limited. A. Ghaisas: Full / Part-time employment: Datar Cancer Genetics Limited. R. Patil: Full / Part-time employment: Datar Cancer Genetics Limited. H. Singh: Full / Parttime employment: Datar Cancer Genetics Limited. A. Srinivasan: Full / Part-time employment: Datar Cancer Genetics Limited. S. Apurwa: Full / Part-time employment: Datar Cancer Genetics Limited. R. Datar: Leadership role, Shareholder / Stockholder / Stock options, Officer / Board of Directors: Datar Cancer Genetics Limited. All other authors have declared no conflicts of interest.
112P
A Bayesian framework to quantify survival uncertainty
H. Loya1, D. Anand1, P. Poduval1, N. Kumar2, A. Sethi2 Department of Electrical Engineering, Indian Institute of Technology Bombay, Mumbai, India, 2Department of Pathology, University of Illinois at Chicago, Chicago, IL, USA
1
Background: Existing survival models do not model uncertainty in personalized (perpatient) prediction of survival, which is useful for treatment planning of cancers. Moreover, their restrictive modeling assumptions limit their accuracy of personalized survival estimation. For example, the Cox proportional hazard (CPH) model assumes a constant effect of each input feature on survival rate over time along with a linear combination of features. The multi-task logistic regression (MTLR) allows a smooth timevarying impact of each feature without relaxing the linearity assumption. Recently,
Volume 30 | Supplement 7 | November 2019
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_7/mdz413.116/5613253 by guest on 30 December 2019
109P
Annals of Oncology
abstracts
Annals of Oncology
Table: 112P Comparison of models Methods
C-index mean (std. dev.)
IBS mean (std. dev.)
CPH MTLR N-MTLR B-MTLR BN-MTLR
0.67 (0.10) 0.68 (0.06) 0.68 (0.02) 0.69 (0.05) 0.70 (0.05)
0.20 (0.07) 0.21 (0.06) 0.16 (0.04) 0.14 (0.02) 0.10 (0.01)
Conclusions: The proposed Bayesian extensions of survival models give better mean personalized accuracy and allow computation of personalized uncertainty scores, which will pave the way for more informative models for treatment planning of cancers. Legal entity responsible for the study: The authors. Funding: Has not received any funding. Disclosure: All authors have declared no conflicts of interest.
114P
S. Shooter, J. Czarnecki, S. Nik-Zainal Medical Genetics, University of Cambridge - MRC Cancer Unit, Cambridge, UK Background: The somatic mutations found through sequencing a whole tumour genome are the aggregate outcome of one or multiple mutational processes. Each of these processes leaves a characteristic imprint, or mutational signature, on the genome. The final mutational profile is determined by the pattern of exposure of each of these mutational processes. Deconstructing the signature composition of a tumour’s mutational profile provides insights into the biological processes that drive it. The initial research in this field extracted 5 substitution signatures from 21 breast cancer samples, but research since has extracted signatures from ever expanding datasets of whole tumour genomes and experimental mutagenesis systems. Signatures have also been developed for different mutation types. The field has great potential to aid in the diagnosis and treatment of cancer patients. Methods: We developed a website using the ReactJS framework supported by a distributed compute cluster in OpenStack. Results: Here we present Signal, a new website enabling exploration of the latest research in the field of mutational signatures. With Signal we intend to inform the scientific community of the power of mutational signatures and to provide the results of our research in a dynamic, easy-to-use interface. This includes tissue-specific substitution and rearrangement signatures extracted from the PanCan dataset, in addition to signatures extracted from cell-lines exposed to environmental mutagens and geneknockouts. Users can upload their own variant calls for analysis by our pipeline, which will filter out kataegis regions, construct a substitution profile and detect the presence of any of our gold-standard substitution signatures. This signature fitting algorithm leverages the organ-specificity of our new signature extraction framework, leading to more accurate signature predictions than previously possible. Conclusions: We intend for Signal to be the home page of mutational signatures which will be kept at the forefront of research with the most recent published data and analysis methodologies. Legal entity responsible for the study: The authors. Funding: CRUK. Disclosure: All authors have declared no conflicts of interest.
115P 113P
Copy number variations in critical cell-signalling genes with potential targeted therapeutic application
E. Nikolova, V. Mitev, A. Todorova Medical Chemistry and Biochemistry, Medical University Sofia, Sofia, Bulgaria Background: Brain tumours are among “hard-to treat” world malignancy with the median survival time of 15 months for the most aggressive form - Glioblastoma multiforme (GBM). It is extremely complex and heterogeneous brain cancer, characterized by rapid progression, infiltration and therapeutic resistance. Currently, many trials for target therapy are focused on understanding the molecular heterogeneity of GBM and designing molecules that inhibit most common genetic alterations. Methods: In the present study, somatic copy number variations (CNVs) in critical signalling genes suitable for future target therapy are examined among Bulgarian patients. Fresh brain biopsies were collected from patients with primary and secondary GBM. Genomic DNA was isolated and CNVs were analysed by Multiplex ligation – dependent probe amplification (MLPA). Results: CDKN2A deletion is the most frequent genetic alteration (78%) among Bulgarian patients with GBM. The loss of this gene is a progression-associated genetic marker. Nearly 60% of GBM patients showed EGFR amplification. It is a prognostic and predictive marker and some patients with this aberration are suggested to respond better to EGFR inhibitors in combination therapy. PTEN deletion is another common event with the incidence rate of 33% and its loss corelates with the choice of mTOR and MEK inhibitors, as well as, the resistance to some EGFR inhibitors. TP53 deletion was found in only 11% of the analysed patients which is a positive marker and the presence of wild-type TP53 is targeted by a MDM2 inhibitors. In addition, MDM2 amplification was observed in one patient and this event also serve as a potential target by MDM2 inhibitors.Amplification in PDGFRA is observed in one patient. Depending on the status of other receptor tyrosine kinases, it could be targeted by PDGFR inhibitor and is also suggested to be a poor prognostic marker. Conclusions: Personalized treatment, customized for an individual patient genetics, has the potential to improve the therapy response. CNVs data could be useful for designing more effective personalized therapy. This is a small study on CNVs with potentially targeted therapeutic application in GBM for Bulgaria and recruitment of more patients is in progress. Legal entity responsible for the study: The authors. Funding: Medical University Sofia, Grant number D-127/2019. Disclosure: All authors have declared no conflicts of interest.
Volume 30 | Supplement 7 | November 2019
Signal: The home page of mutational signatures
Detection of actionable variants in various cancer types reveals value of whole-genome sequencing over in-silico whole-exome and hotspot panel sequencing
K.P. Ramarao-Milne, A-M. Patch, K. Nones, R. Koufariotis, F. Newell, V.R. Addala, O. Kondrashova, P. Mukhopadhyay, S.H. Kazakoff, V. Lakis, O. Holmes, C. Leonard, S. Wood, C. Xu, J.V. Pearson, G. Hollway, N. Waddell Medical Genomics, QIMR Berghofer Medical Research Institute, Brisbane, Australia Background: Declining costs of next-generation sequencing technologies has led to their more widespread use for the identification of targetable mutations. Nonetheless, while multigene panels are being assimilated into clinical practice, the use of wholeexome and whole-genome sequencing (WES and WGS) remains limited, as its added value is unclear in the clinic. Methods: In our study, we have used the Cancer Genome Interpreter (CGI) to identify the actionable variants that are detected by WGS and compared this to in-silico downsampled regions of WES and a hotspot panel. Results: We show that within some tumour types, WGS is invaluable in identifying actionable variants that involve copy number defects and structural aberrations. Moreover, WGS successfully identifies many resistance biomarkers that would have otherwise been missed by WES and panels, particularly in the breast cancer dataset studied. WGS is more valuable of the detection of actionable variants in some tumour types, whereas hotspot panels perform well for other tumour types. Additionally, we show that a tumour type-agnostic approach to selecting genomic biomarker-based drug allocation increases the number of possible FDA-approved and clinical trialbased drug prescriptions for any given genomic biomarker in the patient cohorts studied. Conclusions: Taken together, we show that WGS has the potential to impact patient care by identifying more targetable mutations and therefore expand drug options available for the patient. Furthermore, we illustrate the potential of tumour type-agnostic drug repurposing in nine cancer datasets. Legal entity responsible for the study: QIMR Berghofer Medical Research Institute. Funding: QIMR Berghofer Medical Research Institute. Disclosure: All authors have declared no conflicts of interest.
doi:10.1093/annonc/mdz413 | vii33
Downloaded from https://academic.oup.com/annonc/article-abstract/30/Supplement_7/mdz413.116/5613253 by guest on 30 December 2019
neural-MTLR (N-MTLR) was proposed to relax the linearity assumption as well. We propose Bayesian (B-MTLR) and Bayesian neural network (BN-MTLR) frameworks for predicting survival that are more accurate than the previous methods and compute uncertainty for each personalized prediction. We demonstrate that including more granular information, such as PAM50 gene expressions instead of the PAM50 labels, both increases accuracy and decreases uncertainty in personalized survival prediction. Methods: Using four-fold cross validation on the TCGA-BRCA we compared the accuracy of personalized prediction (C-index and IBS) for CPH, MTLR, N-MTLR, BMTLR, and BN-MTLR using PAM50 labels and clinical variables. We also compared BN-MTLR models trained on PAM50 gene-set and clinical variables for decrease in uncertainty due to more granular inputs. Results: The table below shows that for PAM50 labels and clinical variables the C-index is higher and the IBS is lower for B-MTLR and BN-MTLR compared to the existing techniques. Additionally, BN-MTLR gave a mean uncertainty of 0.12 using PAM50 gene expression as opposed to 0.22 using PAM50 labels across all patients.