RESUMO
Objective: Circulating tumor DNA (ctDNA) is increasingly being used as a potential prognosis biomarker in patients of breast cancer. This review aims to assess the clinical value of ctDNA in outcome prediction in breast cancer patients throughout the whole treatment cycle. Methods: PubMed, Web of Science, Embase, Cochrane Library, Scopus, and clinical trials.gov were searched from January 2016 to May 2022. Conference abstracts published in last three years were also included. The following search terms were used: ctDNA OR circulating tumor DNA AND breast cancer OR breast carcinoma. Only studies written in English languages were included. The following pre-specified criteria should be met for inclusion: (1) observational studies (prospective or retrospective), randomized control trials, case-control studies and case series studies; (2) patients with breast cancer; (3) ctDNA measurement; (4) clinical outcome data such as objective response rate (ORR), pathological complete response (pCR), relapse-free survival (RFS), overall survival (OS), and so on. The random-effect model was preferred considering the potential heterogeneity across studies. The primary outcomes included postoperative short-term outcomes (ORR and pCR) and postoperative long-term outcomes (RFS, OS, and relapse). Secondary outcomes focused on ctDNA detection rate. Results: A total of 30 studies, comprising of 19 cohort studies, 2 case-control studies and 9 case series studies were included. The baseline ctDNA was significantly negatively associated with ORR outcome (Relative Risk [RR] = 0.65, 95% confidence interval [CI]: 0.50-0.83), with lower ORR in the ctDNA-positive group than ctDNA-negative group. ctDNA during neoadjuvant therapy (NAT) treatment was significantly associated with pCR outcomes (Odds Ratio [OR] = 0.15, 95% CI: 0.04-0.54). The strong association between ctDNA and RFS or relapse outcome was significant across the whole treatment period, especially after the surgery (RFS: Hazard Ratio [HR] = 6.74, 95% CI: 3.73-12.17; relapse outcome: RR = 7.11, 95% CI: 3.05-16.53), although there was heterogeneity in these results. Pre-operative and post-operative ctDNA measurements were significantly associated with OS outcomes (pre-operative: HR = 2.03, 95% CI: 1.12-3.70; post-operative: HR = 6.03, 95% CI: 1.31-27.78). Conclusions: In this review, ctDNA measurements at different timepoints are correlated with evaluation indexes at different periods after treatment. The ctDNA can be used as an early potential postoperative prognosis biomarker in breast cancer, and also as a reference index to evaluate the therapeutic effect at different stages.
RESUMO
The prediction of prognostic outcome is critical for the development of efficient cancer therapeutics and potential personalized medicine. However, due to the heterogeneity and diversity of multimodal data of cancer, data integration and feature selection remain a challenge for prognostic outcome prediction. We proposed a deep learning method with generative adversarial network based on sequential channel-spatial attention modules (CSAM-GAN), a multimodal data integration and feature selection approach, for accomplishing prognostic stratification tasks in cancer. Sequential channel-spatial attention modules equipped with an encoder-decoder are applied for the input features of multimodal data to accurately refine selected features. A discriminator network was proposed to make the generator and discriminator learning in an adversarial way to accurately describe the complex heterogeneous information of multiple modal data. We conducted extensive experiments with various feature selection and classification methods and confirmed that the CSAM-GAN via the multilayer deep neural network (DNN) classifier outperformed these baseline methods on two different multimodal data sets with miRNA expression, mRNA expression and histopathological image data: lower-grade glioma and kidney renal clear cell carcinoma. The CSAM-GAN via the multilayer DNN classifier bridges the gap between heterogenous multimodal data and prognostic outcome prediction.
Assuntos
Carcinoma de Células Renais , Glioma , Neoplasias Renais , MicroRNAs , Humanos , PrognósticoRESUMO
Although great progress has been made in prognostic outcome prediction, small sample size remains a challenge in obtaining accurate and robust classifiers. We proposed the Rescaled linear square Regression based Least Squares Learning (RRLSL), a jointly developed semi-supervised feature selection and classifier, for predicting prognostic outcome of cancer patients. RRLSL used the least square regression to identify the scale factors and then rank the features in available multiple types of molecular data. We applied the unlabeled multiple molecular data in conjunction with the labeled data to develop a similarity graph. RRLSL produced the constraint with kernel functions to bridge the gap between label information and geometry information from messenger RNA and microRNA expression profiling. Importantly, this semi-supervised model proposed the least squares learning with L2 regularization to develop a semi-supervised classifier. RRLSL suggested the performance improvement in the prognostic outcome prediction and successfully discriminated between the recurrent patients and non-recurrent ones. We also demonstrated that RRLSL improved the accuracy and Area Under the Precision Recall Curve (AUPRC) as compared to the baseline semi-supervised methods. RRLSL is available for a stand-alone software package (https://github.com/ShiMGLab/RRLSL). A short abstract We proposed the Rescaled linear square Regression based Least Squares Learning (RRLSL), a jointly developed semi-supervised feature selection and classifier, for predicting prognostic outcome of cancer patients. RRLSL used the least square regression to identify the scale factors to rank the features in available multiple types of molecular data. RRLSL produced the constraint with kernel functions to bridge the gap between label information and geometry information from messenger RNA and microRNA expression profiling. Importantly, this semi-supervised model proposed the least squares learning with L2 regularization to develop the semi-supervised classifier. RRLSL suggested the performance improvement in the prognostic outcome prediction and successfully discriminated between the recurrent patients and non-recurrent ones.