Hierarchical Active Learning With Qualitative Feedback on Regions.

Learning classification models in practice usually requires numerous labeled data for training. However, instance-based annotation can be inefficient for humans to perform. In this article, we propose and study a new type of human supervision that is fast to perform and useful for model learning. Instead of labeling individual instances, humans provide supervision to data regions, which are subspaces of the input data space, representing subpopulations of data. Since labeling now is performed on a region level, 0/1 labeling becomes imprecise. Thus, we design the region label to be a qualitative assessment of the class proportion, which coarsely preserves the labeling precision but is also easy for humans to do. To identify informative regions for labeling and learning, we further devise a hierarchical active learning process that recursively constructs a region hierarchy. This process is semisupervised in the sense that it is driven by both active learning strategies and human expertise, where humans can provide discriminative features. To evaluate our framework, we conducted extensive experiments on nine datasets as well as a real user study on a survival analysis of colorectal cancer patients. The results have clearly demonstrated the superiority of our region-based active learning framework against many instance-based active learning methods.

Development of an efficient iterative genome editing method in Bacillus subtilis using the CRISPR-AsCpf1 system.

Bacillus subtilis is a useful chassis in the fields of synthetic biology and metabolic engineering for chemical production. Here, we constructed CRISPR-AsCpf1-based expression plasmids with the temperature-sensitive replicon for iterative genome editing in B. subtilis. This method allowed gene insertion and large genomic deletion with an editing efficiency of up 80%-100% and rapid plasmid curing to facilitate the iterative genome editing in B. subtilis 168. Using the customized CRISPR-AsCpf1 system, we successfully and efficiently implemented the related gene editing in B. subtilis 168 for hyaluronic acid (HA) biosynthesis, HA synthase gene (hasA) insertion, UDP-glucose-dehydrogenase gene (tuaD) insertion, and eps gene cluster (epsA-O) deletion. The heterologous production of HA was realized by the engineered strain with a yield of 1.39 g/L. These results support the finding that the CRISPR-AsCpf1 system is highly efficient in bacteria genome editing and provide valuable guidance and essential references for genome engineering in B. subtilis using the CRISPR-AsCpf1 system.

Self-assembled ternary hybrid nanodrugs for overcoming tumor resistance and metastasis.

Traditional chemotherapy exhibits a certain therapeutic effect toward malignant cancer, but easily induce tumor multidrug resistance (MDR), thereby resulting in the progress of tumor recurrence or metastasis. In this work, we deigned ternary hybrid nanodrugs (PEI/DOX@CXB-NPs) to simultaneously combat against tumor MDR and metastasis. In vitro results demonstrate this hybrid nanodrugs could efficiently increase cellular uptake at pH 6.8 by the charge reversal, break lysosomal sequestration by the proton sponge effect and trigger drugs release by intracellular GSH, eventually leading to higher drugs accumulation and cell-killing in drug-sensitive/resistant cells. In vivo evaluation revealed that this nanodrugs could significantly inhibit MDR tumor growth and simultaneously prevent A549 tumor liver/lung metastasis owing to the specifically drugs accumulation. Mechanism studies further verified that hybrid nanodrugs were capable of down-regulating the expression of MDR or metastasis-associated proteins, lead to the enhanced anti-MDR and anti-metastasis effect. As a result, the multiple combination strategy provided an option for effective cancer treatment, which could be potentially extended to other therapeutic agents or further use in clinical test.

pH-sensitive and tumor-targeting nanogels based on ortho ester-modified PEG for improving the in vivo anti-tumor efficiency of doxorubicin.

In this study, we aim to develop the pH-sensitive and tumor-targeting nanogels based on the co-polymerization of three terminal allyl-functionalized components, including ortho ester-conjugated mPEG (mPEG-MOE), ortho ester crosslinker (OEAM) and phenylboronic acid (APBA). The hybrid nanogels displayed a typical spherical structure with a diameter around 200 nm observed by dynamic light scattering (DLS) and scanning electron microscopy (SEM). The prepared nanogels possessed a good stability in neutral conditions, while displayed pH-triggered drug release profiles. Furthermore, in vitro study of cellular uptake and cytotoxicity indicated that the nanogels possessed the highest drug accumulation and cytotoxicity against EMT6 cells. In vivo antitumor examination suggested that these nanogels brought out excellent efficacy in enhancing drug concentration, restraining tumor growth, and prolonged the survival time of tumor-bearing mice. Thus, the prepared multi-functional nanogels possess great potentials for drug delivery in tumor treatment.

Hybrid nanoparticles based on ortho ester-modified pluronic L61 and chitosan for efficient doxorubicin delivery.

Tumor intrinsic or acquired multidrug resistance (MDR) is still one of the major obstacles to the success of nanomedicine. To address this, the pH-sensitive nanoparticles (L61-OE-CS) with MDR-reversal ability were prepared by the crosslinking between acid-labile ortho-ester-modified pluronic (L61-OE) and chitosan (CS) for efficient doxorubicin (DOX) delivery. The size and micromorphology of the prepared nanoparticles were observed by dynamic light scanning and scanning electron microscopy and the nanoparticles displayed a uniform spherical shape with a diameter around 200 nm. The pH-triggered morphology change of the nanoparticles was also observed by scanning electron microscope. Drug release profiles under different pH values showed that DOX release amount within 72 h reached 16% (pH 7.4) and 76.5% (pH 5.0), respectively. In vitro cellular uptake and MTT assay demonstrated that the ortho ester and pluronic-based nanoparticles had higher cytotoxicity than non-sensitive nanoparticles. In vivo antitumor experiments also proved the superiority of the dual-functional nanoparticles, and the tumor growth inhibition rate (TGI) on day 14 was higher than 80%. Therefore, L61-OE-CS nanoparticles have great potential to be used as drug carriers in anticancer therapy.

Dynamic precise dual-drug-backboned nano-prodrugs for selective chemotherapy.

To achieve an ideal drug delivery platform with precise composition and high tumor selectivity, the PEGylated dual-drug backboned prodrug was synthesized via the copolymerization between diamine monomer of ortho ester and cisplatin- demethylcantharidin conjugate (Pt(IV)-1), and then terminated by mPEG550-active ester. The amphipathic prodrug could self-assemble into nano-prodrugs, which endowed the precise structure and high drug loading. Moreover, the nano-prodrugs exhibited physicochemical stability at physiological pH (7.4) for stable blood circulation, DePEGylation and dynamic size change for selective tumor accumulation and enhanced cellular internalization at tumoral extracellular pH (6.8), and efficient drug release for synergetic apoptosis and cytotoxicity at tumoral intracellular pH (5.0)/glutathione. Thus, the precise dual-drug backboned nano-prodrugs with detachable PEGylation, dynamic size change and efficient drug release could be potentially translated for clinically selective cancer treatment. STATEMENT OF SIGNIFICANCE: Few nanomedicines have been clinically used for cancer treatment and little progress has been made in the last decades due to the unprecise composition and unsatisfactory tumor selectivity. Herein, the PEGylated dual-drug backboned nano-prodrugs were successfully constructed by rational design and endowed the defined structure, precise drug ratio, extraordinary high drug loading and reduction/pH dual sensitivity. The nano-prodrugs further exhibited the stable storage and blood circulation through PEGylation and low critical micelle concentration, enhanced tumor accumulation and cellular uptake via extracellular DePEGylation and dynamic size translation, and synergetic cytotoxicity via intracellular efficient drug release, respectively. This study can open a new avenue for easy industrial manufacture and quality control, and highly selective chemotherapy appealing for clinical translation. .

Self-assembled 5-fluorouracil-cinnamaldehyde nanodrugs for greatly improved chemotherapy in vivo.

The preferred cancer treatment is to achieve a high therapeutic effect as well as reduce side effects. In this study, we developed carrier-free nano drugs based on 5-fluorouracil (5FU) and cinnamaldehyde (CA) to meet the above goals. Two model drugs were spliced by acetal linkage and ester bond, which could self-assemble into nano drug particles (5FU-CA NPs) with a size of ∼170 nm. In vitro cell experiments showed 5FU-CA NPs were efficiently internalized by HepG2 cells. They then quickly exerted dual drug activities by the cleavage of acetal and ester bond, resulting in enhanced cell-killing efficacy and apoptosis. Synergistic mechanisms were achieved via the anti-metabolic effects mediated by 5FU-COOH and the oxidative damage induced by CA. In vivo anti-tumor evaluation further indicated that 5FU-CA NPs had higher tumor growth inhibition than 5FU-COOH/CA mixture (5FU-COOH + CA) and exhibited lower systemic toxicity under the same reducing dose of each drug. Overall, this is a successful synergistic anti-tumor attempt through rational self-assembly of drugs with different mechanisms and it can be extrapolated to other agents.

pH-sensitive deoxycholic acid dimer for improving doxorubicin delivery and antitumor activity in vivso.

To develop simple and effective nano-drug delivery systems remains a major challenge in cancer treatment. Herein, we synthesized an ortho ester-linked deoxycholic acid dimer (DCA-OE), which could effectively self-assemble with doxorubicin (DOX) to form stable nanoparticles (DCA-OE/DOX NPs) by a single emulsion method. DCA-based nanoparticles had a desirable size (∼200 nm), morphology (spherical shape), and high drug encapsulation (drug loading content of ∼18.0 %, drug loading efficiency of ∼77.6 %). DCA-OE could improve the stability and solubility of DOX in physiological environment, while pH-sensitive ortho ester linkage endowed the ability to release DOX quickly in cancer cells. In vitro cytotoxicity and apoptosis verified drug-loaded dimer nanoparticles had similar toxicity with free DOX. Besides, these particles could efficiently accumulate and penetrate into human liver carcinoma cell line (HepG2) multicellular spheroids, thus resulting in enhanced antitumor effect. In vivo tests further exhibited that DCA-OE/DOX NPs had lower systemic toxicity and higher tumor inhibition effect, and its tumor inhibition rate was 84.1 %, which was far more than free DOX (49.3 %). Therefore, the strategy to link functional small molecules with ortho ester has great potentials in specific delivery of anticancer drugs.

Oxygen-producing catalase-based prodrug nanoparticles overcoming resistance in hypoxia-mediated chemo-photodynamic therapy.

Extreme hypoxia inside solid tumors is the primary barrier against the advance of chemotherapy and photodynamic therapy (PDT). To address this problem, a hybrid nano-enzyme prodrug system was developed to alleviate hypoxia as well as simultaneously sensitize chemo-photodynamic therapy. Lactobionic acid (LA) and doxorubicin (DOX) precursor (cis-aconitic anhydride-linked doxorubicin, CAD) were pre-conjugated onto the side chain of catalase (CAT), then co-assembled with chlorin e6 (Ce6) to form LA-CAT-CAD@Ce6 nanoparticles (LCC@Ce6-NPs). LA as the active-targeting ligand increased cellular internalization, CAD as the pH-sensitive component triggered rapid drug release, Ce6 as the photosensitizer induced reactive oxygen species (ROS) generation, and CAT decomposed intracellular H2O2 to produce oxygen in situ. Oxygen production efficiently decreased the expression of hypoxia-inducible factor-1α (HIF-1α) and P-glycoprotein (P-gp), which enhanced chemotherapy efficiency. In addition, sufficient oxygen further amplified PDT-mediated cell-killing and apoptosis in hypoxic tumor. In vivo studies showed that combined chemo-photodynamic therapy by LCC@Ce6-NPs led to the most effective inhibition of tumor growth (TGI>90%), and even partially ablated tumor. Thus, this nano-enzyme prodrug platform can be a potentially effective treatment in clinical cancer therapy, and married to other therapeutic agents. STATEMENT OF SIGNIFICANCE: Hypoxia in solid tumors seriously impedes the efficacy of chemotherapy or photodynamic therapy. Herein, we designed hybrid nano-enzyme prodrug particles to improve hypoxia-mediated limitations on cancer therapy. Lactobionic acid (LA) as the hydrophilic outer layer of particles increased cellular uptake by receptor-mediated endocytosis, and cis-aconitic anhydride-linked doxorubicin (CAD) as the pH sensitive component inside particles efficiently triggered DOX and Ce6 release. More importantly, catalase (CAT) as the backbone of particles was capable of greatly relieving tumor hypoxia through catalyzing the decomposition of H2O2 in situ. Oxygen re-generation not only prevented hypoxia-mediated chemo-resistance, but also amplified PDT-induced ROS cell-killing ability. As a result, the multiple combination action of this nano-system could simultaneously sensitize chemo-photodynamic therapy, thus significantly enhancing tumor therapy.

Active Learning of Multi-class Classification Models from Ordered Class Sets.

In this paper, we study the problem of learning multi-class classification models from a limited set of labeled examples obtained from human annotator. We propose a new machine learning framework that learns multi-class classification models from ordered class sets the annotator may use to express not only her top class choice but also other competing classes still under consideration. Such ordered sets of competing classes are common, for example, in various diagnostic tasks. In this paper, we first develop strategies for learning multi-class classification models from examples associated with ordered class set information. After that we develop an active learning strategy that considers such a feedback. We evaluate the benefit of the framework on multiple datasets. We show that class-order feedback and active learning can reduce the annotation cost both individually and jointly.

Active Learning of Multi-Class Classifiers with Auxiliary Probabilistic Information.

Our ability to learn accurate classification models from data is often limited by the number of available data instances. This limitation is of particular concern when data instances need to be labeled by humans and when the labeling process carries a significant cost. Recent years witnessed increased research interest in developing methods capable of learning models from a smaller number of examples. One such direction is active learning. Another, more recent direction showing a great promise utilizes auxiliary probabilistic information in addition to class labels. However, this direction has been applied and tested only in binary classification settings. In this work we first develop a multi-class variant of the auxiliary probabilistic approach, and after that embed it within an active learning framework, effectively combining two strategies for reducing the dependency of multi-class classification learning on the number of labeled examples. We demonstrate the effectiveness of our new approach on both simulated and real-world datasets.

Active Learning of Classification Models with Likert-Scale Feedback.

Annotation of classification data by humans can be a time-consuming and tedious process. Finding ways of reducing the annotation effort is critical for building the classification models in practice and for applying them to a variety of classification tasks. In this paper, we develop a new active learning framework that combines two strategies to reduce the annotation effort. First, it relies on label uncertainty information obtained from the human in terms of the Likert-scale feedback. Second, it uses active learning to annotate examples with the greatest expected change. We propose a Bayesian approach to calculate the expectation and an incremental SVM solver to reduce the time complexity of the solvers. We show the combination of our active learning strategy and the Likert-scale feedback can learn classification models more rapidly and with a smaller number of labeled instances than methods that rely on either Likert-scale labels or active learning alone.

Efficient Learning of Classification Models from Soft-label Information by Binning and Ranking.

Construction of classification models from data in practice often requires additional human effort to annotate (label) observed data instances. However, this annotation effort may often be too costly and only a limited number of data instances may be feasibly labeled. The challenge is to find methods that let us reduce the number of the labeled instances but at the same time preserve the quality of the learned models. In this paper we study the idea of learning classification from soft label information in which each instance is associated with a soft-label further refining its class label. One caveat of applying this idea is that soft-labels based on human assessment are often noisy. To address this problem, we develop and test a new classification model learning algorithm that relies on soft-label binning to limit the effect of soft-label noise. We show this approach is able to learn classification models more rapidly and with a smaller number of labeled instances than (1) existing soft label learning methods, as well as, (2) methods that learn from class-label information.