Abstract 4298: Machine learning-based model for survival prediction after immunotherapy in patients with solid tumor

Background: Immune checkpoint inhibitors (ICIs) have led to a paradigm shift in solid tumors treatment. However, not all patients respond favorably to these drugs, highlighting the need for reliable predictions to achieve more personalized care and better management. This study aimed to create and validated ML model to predict survival in solid tumors patients receiving ICIs.

Methods: We obtained clinical and genomic data from cBio Cancer Genomics Portal. Data were randomly divided into a training set (80%) and a validation set (20%). light gradient boosting algorithm was trained to predict patients’ survival at different time points.

Results: We identified 1660 patients with a median survival of 18 months. LGB yielded AUCs of 67.91% at 1 year, 79.89.6% at 2 years, and 79.75% at 3 years, respectively. The most important predictors that influenced the performance of the model in predicting 3-year-survival were: Age (22.89%), tumor mutational burden (18.24%), and tumor purity (13.23%). Moreover, multivariate analysis was performed and drug type was identified as an independent prognostic indicator (P< .001). So, a Subgroup analysis was done and the OS rates were: 98.57%, 75.55%, 33.84% in patients who received CTLA-4, 98.57%, 75.55%, 33.84% with PD-1/PD-L1, and 97.99%, 9.94%, 4.62% with combo treatment, at 1-, 3-, and 5 years, respectively.

Conclusion: Our ML-based model that integrates both clinical and genomic data is an improved tool for survival prediction, enabling an accurate risk classification and leading to a more precise decision-making. Moreover, this study highlights the importance of age and tumor microenvironment as the main contributors in making survival prediction in patients receiving ICIs. Table. Performance in survival prediction among training and testing sets

Table 1. Survival Average AUC Average Accuracy Survival Duration Positive Predictive Value (Precision) Sensitivity (Recall) 1-year 67.91% 63.18% <12 months 58% 54% >=12 months 67% 70% 2-years 79.89% 78.70% <24 months 86% 84% >=24 months 62% 65% 3-years 79.75% 86.02% <36 months 95% 89% >=36 months 45% 65%

Citation Format: Salma Y. Fala, Mohamed Osman. Machine learning-based model for survival prediction after immunotherapy in patients with solid tumor. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4298.

Structured crowdsourcing enables convolutional segmentation of histology images

MOTIVATION

While deep-learning algorithms have demonstrated outstanding performance in semantic image segmentation tasks, large annotation datasets are needed to create accurate models. Annotation of histology images is challenging due to the effort and experience required to carefully delineate tissue structures, and difficulties related to sharing and markup of whole-slide images.

RESULTS

We recruited 25 participants, ranging in experience from senior pathologists to medical students, to delineate tissue regions in 151 breast cancer slides using the Digital Slide Archive. Inter-participant discordance was systematically evaluated, revealing low discordance for tumor and stroma, and higher discordance for more subjectively defined or rare tissue classes. Feedback provided by senior participants enabled the generation and curation of 20 000+ annotated tissue regions. Fully convolutional networks trained using these annotations were highly accurate (mean AUC=0.945), and the scale of annotation data provided notable improvements in image classification accuracy.

AVAILABILITY AND IMPLEMENTATION

Dataset is freely available at: https://goo.gl/cNM4EL.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Reporting quality in systematic reviews of in vitro studies: a systematic review

Background: Systematic reviews (SRs) and/or meta-analyses of in vitro research have an important role in establishing the foundation for clinical studies. In this study, we aimed to evaluate the reporting quality of SRs of in vitro studies using the PRISMA checklist.Method: Four databases were searched including PubMed, Virtual Health Library (VHL), Web of Science (ISI) and Scopus. The search was limited from 2006 to 2016 to include all SRs and/or meta-analyses (MAs) of pure in vitro studies. The evaluation of reporting quality was done using the PRISMA checklist.Results: Out of 7702 search results, 65 SRs were included and evaluated with the PRISMA checklist. Overall, the mean overall quality score of reported items of the PRISMA checklist was 68%. We have noticed an increasing pattern in the numbers of published SRs of in vitro studies over the last 10 years. In contrast, the reporting quality was not significantly improved over the same period (p = .363). There was a positive but not significant correlation between the overall quality score and the journal impact factor of the included studies.Conclusions: The adherence of SRs of in vitro studies to the PRISMA guidelines was poor. Therefore, we believe that using reporting guidelines and journals paying attention to this fact will improve the quality of SRs of in vitro studies.

Oncogenic long noncoding RNA MALAT1 and HCV-related hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related mortality worldwide. The oncogenic function of the long non-coding RNA; metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) in HCC remains unclear. We aimed to evaluate MALAT1 serum expression profile in HCC and explore its relation to the clinicopathological features. Quantitative Real Time-Polymerase Chain Reaction was applied in 70 cohorts (30 HCC, 20 HCV, 20 controls). Further meta-analysis of clinical studies and in vitro validated experiments was employed. Serum MALAT1 showed area under the curve of 0.79 and 0.70 to distinguish patients with cancer from normal and cirrhotic individuals at fold change of 1.0 and 1.26, respectively. Expression level was significantly higher in males (P <0.001) and patients with massive ascites (P = 0.005). Correlation analysis showed positive correlation of MALAT1 with total bilirubin (r = 0.456, P <0.001) and AST (r = 0.280, P = 0.019), and negative correlation with the hemoglobin level (r = 0.312, P = 0.009). Meta-analysis showed that the over-expressed MALAT1 was linked to tumor number [Cohen's d = 0.450, 95% CI (0.21 to 0.68)], clinical stage [Cohen's d = 0.048, 95% CI (-0.83 to 0.74)], and AFP level [Cohen's d = 0.354, 95% CI (0.1 to 0.57)]. In silico data analysis and systematic review confirmed MALAT1 oncogenic function in cancer development and progression. In conclusion, circulatory MALAT1 might represent a putative non-invasive prognostic biomarker indicating worse liver failure score in HCV-related HCC patients with traditional markers. Large-scale verification is warranted in future studies.