Search

Accurate prediction of cancer stage is important in that it enables more appropriate treatment for patients with cancer. Many measures or methods have been proposed for more accurate prediction of cancer stage, but recently, machine learning, especially deep learning-based methods have been receiving increasing attention, mostly owing to their good prediction accuracy in many applications. Machine learning methods can be applied to high throughput DNA mutation or RNA expression data to predict cancer stage. However, because the number of genes or markers generally exceeds 10,000, a considerable number of data samples is required to guarantee high prediction accuracy. To solve this problem of a small number of clinical samples, we used a Generative Adversarial Networks (GANs) to augment the samples. Because GANs are not effective with whole genes, we first selected significant genes using DNA mutation data and random forest feature ranking. Next, RNA expression data for selected genes were expanded using GANs. We compared the classification accuracies using original dataset and expanded datasets generated by proposed and existing methods, using random forest, Deep Neural Networks (DNNs), and 1-Dimensional Convolutional Neural Networks (1DCNN). When using the 1DCNN, the F1 score of GAN5 (a 5-fold increase in data) was improved by 39% in relation to the original data. Moreover, the results using only 30% of the data were better than those using all of the data. Our attempt is the first to use GAN for augmentation using numeric data for both DNA and RNA. The augmented datasets obtained using the proposed method demonstrated significantly increased classification accuracy for most cases. By using GAN and 1DCNN in the prediction of cancer stage, we confirmed that good results can be obtained even with small amounts of samples, and it is expected that a great deal of the cost and time required to obtain clinical samples will be reduced. The proposed sample augmentation method could also be applied for other purposes, such as prognostic prediction or cancer classification., Competing Interests: The authors have declared that no competing interests exist.

The better understanding of polymeric nanoparticles as a drug delivery carrier is a decisive factor to get more efficient therapeutic response in vivo. Here, we report the non-invasive imaging of bare polymeric nanoparticles and drug-loaded polymeric nanoparticles to evaluate biodistribution in tumor bearing mice. To make nano-sized drug delivery carrier, glycol chitosan was modified with different degrees of hydrophobic N-acetyl histidine (NAcHis-GC-1, -2, and -3). The biodistribution of polymeric nanoparticles and drug was confirmed by using gamma camera with (131)I-labeled NAcHis-GC and (131)I-labeled doxorubicin (DOX) and by using in vivo live animal imaging with near-infrared fluorescence Cy5.5-labeled NAcHis-GC. Among bare nanoparticles, NAcHis-GC3 (7.8% NAcHis content) showed much higher tumor targeting efficiency than NAcHis-GC1 (3.3% NAcHis content) and NAcHis-GC2 (6.8% NAcHis content). In contrast, for drug-loaded nanoparticles, DOX-NAcHis-GC1 displayed two-fold higher tumor targeting property than DOX-NAcHis-GC3. These data imply that the biodistribution and tumor targeting efficiency between bare and drug-loaded nanoparticles may be greatly different. Therapeutic responses for NAcHis-GC nanoparticles after drug loading were also evaluated. In xenograft animal model, we could find out that DOX-NAcHis-GC1 with higher tumor targeting of DOX has more excellent therapeutic effect than DOX-NAcHis-GC3 and free DOX. These results mean that the hydrophobic core stability might be a critical factor for tumor targeting efficiency of nanoparticles. The present study indicates that by using molecular imaging, we can select more appropriate nanoparticles with the highest tumor targeting properties, leading to exerting more excellent therapeutic results in cancer therapy., (Copyright © 2010 Elsevier B.V. All rights reserved.)

Herein, we evaluated the tumoral low pH targeting characteristics of pH-responsive polymer micelles in cancer targeting therapy. To design the pH-responsive polymeric micelles, hydrophilic methyl ether poly(ethylene glycol) (MPEG) and pH-responsive/biodegradable poly(beta-amino ester) (PAE) were copolymerized using a Michael-type step polymerization, resulting in an MEPG-PAE block copolymer. The amphiphilic MPEG-PAE block copolymer formed polymeric micelles with nano-sized diameter by self-assembly, which showed a sharp pH-dependant micellization/demicellization transition at the tumoral acidic pH value (pH 6.4). For the cancer image and therapy, fluorescence dye, tetramethylrhodamine isothiocyanate (TRITC), or anticancer drug, camptothecin (CPT), was efficiently encapsulated into the pH-responsive polymeric micelles (pH-PMs) by a simple solvent casting method. The TRITC or CPT encapsulated pH-PMs (TRITC-pH-PMs or CPT-pH-PMs) showed rapid release of TRITC or CPT in weakly acidic aqueous (pH 6.4) because they still presented a sharp tumoral acid pH-responsive micellization/demicellization transition. The pH-PMs with 10wt.% of TRITC could deliver substantially more fluorescence dyes to the target tumor tissue in MDA-MB231 human breast tumor-bearing mice, compared to the control polymeric micelles of PEG-poly(l-lactic acid) (PEG-PLLA). Importantly, CPT-pH-PMs exhibited significantly increased therapeutic efficacy with minimum side effects by other tissues in breast tumor-bearing mice, compared to free CPT and CPT encapsulated PEG-PLLA micelles. The tumoral acidic pH-responsive polymeric micelles are highly useful for cancer targeting therapy., (Copyright (c) 2010 Elsevier B.V. All rights reserved.)

We report tumor targeting nanoparticles for optical/MR dual imaging based on self-assembled glycol chitosan to be a potential multimodal imaging probe. To develop an optical/MR dual imaging probe, biocompatible and water-soluble glycol chitosan (M(w) = 50 kDa) were chemically modified with 5beta-cholanic acid (CA), resulting in amphiphilic glycol chitosan-5beta-cholanic acid conjugates (GC-CA). For optical imaging near-infrared fluorescence (NIRF) dye, Cy5.5, was conjugated to GC-CA resulting in Cy5-labeled GC-CA conjugates (Cy5.5-GC-CA). Moreover, in order to chelate gadolinium (Gd(III)) in the Cy5.5-GC-CA conjugates, 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) was directly conjugated in Cy5.5-GC-CA. Finally, the excess GdCl(3) was added to DOTA modified Cy5.5-GC-CA conjugates in distilled water (pH 5.5). The freshly prepared Gd(III) encapsulated Cy5.5-GC-CA conjugates were spontaneously self-assembled into stable Cy5.5 labeled and Gd(III) encapsulated chitosan nanoparticles (Cy5.5-CNP-Gd(III)). The Cy5.5-CNP-Gd(III) was spherical in shape and approximately 350 nm in size. From the cellular experiment, it was demonstrated that Cy5.5-CNP-Gd(III) were efficiently taken up and distributed in cytoplasm (NIRF filter; red). When the Cy5.5-GC-Gd(III) were systemically administrated into the tail vein of tumor-bearing mice, large amounts of nanoparticles were successfully localized within the tumor, which was confirmed by noninvasive near-infrared fluorescence and MR imaging system simultaneously. These results revealed that the dual-modal imaging probe of Cy5.5-CNP-Gd(III) has the potential to be used as an optical/MR dual imaging agent for cancer treatment.

We report the fabrication and characterization of thermally cross-linked superparamagnetic iron oxide nanoparticles (TCL-SPION) and their application to the dual imaging of cancer in vivo. Unlike dextran-coated cross-linked iron oxide nanoparticles, which are prepared by a chemical cross-linking method, TCL-SPION are prepared by a simple, thermal cross-linking method using a Si-OH-containing copolymer. The copolymer, poly(3-(trimethoxysilyl)propyl methacrylate-r-PEG methyl ether methacrylate-r-N-acryloxysuccinimide), was synthesized by radical polymerization and used as a coating material for as-synthesized magnetite (Fe3O4) SPION. The polymer-coated SPION was further heated at 80 degrees C to induce cross-linking between the -Si(OH)3 groups in the polymer chains, which finally generated TCL-SPION bearing a carboxyl group as a surface functional group. The particle size, surface charge, presence of polymer-coating layers, and the extent of thermal cross-linking were characterized and confirmed by various measurements, including dynamic light scattering, Fourier transform infrared spectroscopy, and X-ray photoelectron spectroscopy. The carboxyl TCL-SPION was converted to amine-modified TCL-SPION and then finally to Cy5.5 dye-conjugated TCL-SPION for use in dual (magnetic resonance/optical) in vivo cancer imaging. When the Cy5.5 TCL-SPION was administered to Lewis lung carcinoma tumor allograft mice by intravenous injection, the tumor was unambiguously detected in T2-weighted magnetic resonance images as a 68% signal drop as well as in optical fluorescence images within 4 h, indicating a high level of accumulation of the nanomagnets within the tumor site. In addition, ex vivo fluorescence images of the harvested tumor and other major organs further confirmed the highest accumulation of the Cy5.5 TCL-SPION within the tumor. It is noteworthy that, despite the fact that TCL-SPION does not bear any targeting ligands on its surface, it was highly effective for tumor detection in vivo by dual imaging.