Search for: All records

Multi-instance learning (MIL) handles data that is organized into sets of instances known as bags. Traditionally, MIL is used in the supervised-learning setting for classifying bags which contain any number of instances. However, many traditional MIL algorithms do not scale efficiently to large datasets. In this paper, we present a novel primal–dual multi-instance support vector machine that can operate efficiently on large-scale data. Our method relies on an algorithm derived using a multi-block variation of the alternating direction method of multipliers. The approach presented in this work is able to scale to large-scale data since it avoids iteratively solving quadratic programming problems which are broadly used to optimize MIL algorithms based on SVMs. In addition, we improve our derivation to include an additional optimization designed to avoid solving a least-squares problem in our algorithm, which increases the utility of our approach to handle a large number of features as well as bags. Finally, we derive a kernel extension of our approach to learn nonlinear decision boundaries for enhanced classification capabilities. We apply our approach to both synthetic and real-world multi-instance datasets to illustrate the scalability, promising predictive performance, and interpretability of our proposed method. 

As key mediators of cellular communication, extracellular vesicles (EVs) have been actively explored for diagnostic and therapeutic applications. However, effective methods to functionalize EVs and modulate the interaction between EVs and recipient cells are still lacking. Here we report a facile and universal metabolic tagging technology that can install unique chemical tags (e.g., azido groups) onto EVs. The surface chemical tags enable conjugation of molecules via efficient click chemistry, for the tracking and targeted modulation of EVs. In the context of tumor EV vaccines, we show that the conjugation of toll-like receptor 9 agonists onto EVs enables timely activation of dendritic cells and generation of superior antitumor CD8⁺T cell response. These lead to 80% tumor-free survival against E.G7 lymphoma and 33% tumor-free survival against B16F10 melanoma. Our study yields a universal technology to generate chemically tagged EVs from parent cells, modulate EV-cell interactions, and develop potent EV vaccines.

 

Dendritic cell (DC) vaccine was among the first FDA-approved cancer immunotherapies, but has been limited by the modest cytotoxic T lymphocyte (CTL) response and therapeutic efficacy. Here we report a facile metabolic labeling approach that enables targeted modulation of adoptively transferred DCs for developing enhanced DC vaccines. We show that metabolic glycan labeling can reduce the membrane mobility of DCs, which activates DCs and improves the antigen presentation and subsequent T cell priming property of DCs. Metabolic glycan labeling itself can enhance the antitumor efficacy of DC vaccines. In addition, the cell-surface chemical tags (e.g., azido groups) introduced via metabolic glycan labeling also enable in vivo conjugation of cytokines onto adoptively transferred DCs, which further enhances CTL response and antitumor efficacy. Our DC labeling and targeting technology provides a strategy to improve the therapeutic efficacy of DC vaccines, with minimal interference upon the clinical manufacturing process.

 

Inhibition of overexpressed enzymes is among the most promising approaches for targeted cancer treatment. However, many cancer-expressed enzymes are “nonlethal,” in that the inhibition of the enzymes’ activity is insufficient to kill cancer cells. Conventional antibody-based therapeutics can mediate efficient treatment by targeting extracellular nonlethal targets but can hardly target intracellular enzymes. Herein, we report a cancer targeting and treatment strategy to utilize intracellular nonlethal enzymes through a combination of selective cancer stem-like cell (CSC) labeling and Click chemistry-mediated drug delivery. A de novo designed compound, AAMCHO [N-(3,4,6-triacetyl- N-azidoacetylmannosamine)-cis-2-ethyl-3-formylacrylamideglycoside], selectively labeled cancer CSCs in vitro and in vivo through enzymatic oxidation by intracellular aldehyde dehydrogenase 1A1. Notably, azide labeling is more efficient in identifying tumorigenic cell populations than endogenous markers such as CD44. A dibenzocyclooctyne (DBCO)-toxin conjugate, DBCO-MMAE (Monomethylauristatin E), could next target the labeled CSCs in vivo via bioorthogonal Click reaction to achieve excellent anticancer efficacy against a series of tumor models, including orthotopic xenograft, drug-resistant tumor, and lung metastasis with low toxicity. A 5/7 complete remission was observed after single-cycle treatment of an advanced triple-negative breast cancer xenograft (~500 mm³).