High-quality Affinity Research Articles

Auto-weighted multiple kernel tensor clustering

Multiple kernel subspace clustering (MKSC) has attracted intensive attention since its powerful capability of exploring consensus information by generating a high-quality affinity graph from multiple base kernels. However, the existing MKSC methods still exist the following limitations: (1) they essentially neglect the high-order correlations hidden in different base kernels; and (2) they perform candidate affinity graph learning and consensus affinity graph learning in two separate steps, where suboptimal solution may be obtained. To alleviate these problems, a novel MKSC method, namely auto-weighted multiple kernel tensor clustering (AMKTC), is proposed. Specifically, AMKTC first integrates the consensus affinity graph learning and candidate affinity graph learning into a unified framework, where the optimal goal can be achieved by making these two learning processes negotiate with each other. Further, an auto-weighted fusion scheme with one-step manner is proposed to learn the final consensus affinity graph, where the reasonable weights will be automatically learned for each candidate graph. Finally, the essential high-order correlations between multiple base kernels can be captured by leveraging tensor-singular value decomposition (t-SVD)-based tensor nuclear norm constraint on a 3-order graph tensor. Experiments on seven benchmark datasets with eleven comparison methods demonstrate that our method achieves state-of-the-art clustering performance.

Flow-Cell-Based Technology for Massively Parallel Characterization of Base-Modified DNA Aptamers.

Aptamers incorporating chemically modified bases can achieve superior affinity and specificity compared to natural aptamers, but their characterization remains a labor-intensive, low-throughput task. Here, we describe the "non-natural aptamer array" (N2A2) system, in which a minimally modified Illumina MiSeq instrument is used for the high-throughput generation and characterization of large libraries of base-modified DNA aptamer candidates based on both target binding and specificity. We first demonstrate the capability to screen multiple different base modifications to identify the optimal chemistry for high-affinity target binding. We next use N2A2 to generate aptamers that can maintain excellent specificity even in complex samples, with equally strong target affinity in both buffer and diluted human serum. For both aptamers, affinity was formally calculated with gold-standard binding assays. Given that N2A2 requires only minor mechanical modifications to the MiSeq, we believe that N2A2 offers a broadly accessible tool for generating high-quality affinity reagents for diverse applications.

Effects of data quality and quantity on deep learning for protein-ligand binding affinity prediction

Prediction of protein-ligand binding affinities is crucial for computational drug discovery. A number of deep learning approaches have been developed in recent years to improve the accuracy of such affinity prediction. While the predicting power of these systems have advanced to some degrees depending on the dataset used for model training and testing, the effects of the quality and quantity of the underlying data have not been thoroughly examined. In this study, we employed erroneous datasets and data subsets of different sizes, created from one of the largest databases of experimental binding affinities, to train and evaluate a deep learning system based on convolutional neural networks. Our results show that data quality and quantity do have significant impacts on the prediction performance of trained models. Depending on the variations in data quality and quantity, the performance discrepancies could be comparable to or even larger than those observed among different deep learning approaches. In particular, the presence of proteins in the training data leads to a dramatic increase in prediction accuracy. This implies that continued accumulation of high-quality affinity data, especially for new protein targets, is indispensable for improving deep learning models to better predict protein-ligand binding affinities.

Sensitive Protein Detection Using Site-Specifically Oligonucleotide-Conjugated Nanobodies.

High-quality affinity probes are critical for sensitive and specific protein detection, in particular for detection of protein biomarkers in the early phases of disease development. Proximity extension assays (PEAs) have been used for high-throughput multiplexed protein detection of up to a few thousand different proteins in one or a few microliters of plasma. Clonal affinity reagents can offer advantages over the commonly used polyclonal antibodies (pAbs) in terms of reproducibility and standardization of such assays. Here, we explore nanobodies (Nbs) as an alternative to pAbs as affinity reagents for PEA. We describe an efficient site-specific approach for preparing high-quality oligo-conjugated Nb probes via enzyme coupling using Sortase A (SrtA). The procedure allows convenient removal of unconjugated affinity reagents after conjugation. The purified high-grade Nb probes were used in PEA, and the reactions provided an efficient means to select optimal pairs of binding reagents from a group of affinity reagents. We demonstrate that Nb-based PEA (nano-PEA) for interleukin-6 (IL6) detection can augment assay performance, compared to the use of pAb probes. We identify and validate Nb combinations capable of binding in pairs without competition for IL6 antigen detection by PEA.

Multiple Kernel Clustering with Kernel k-Means Coupled Graph Tensor Learning

Kernel k-means (KKM) and spectral clustering (SC) are two basic methods used for multiple kernel clustering (MKC), which have both been widely used to identify clusters that are non-linearly separable. However, both of them have their own shortcomings: 1) the KKM-based methods usually focus on learning a discrete clustering indicator matrix via a combined consensus kernel, but cannot exploit the high-order affinities of all pre-defined base kernels; and 2) the SC-based methods require a robust and meaningful affinity graph in kernel space as input in order to form clusters with desired clustering structure. In this paper, a novel method, kernel k-means coupled graph tensor (KCGT), is proposed to graciously couple KKM and SC for seizing their merits and evading their demerits simultaneously. In specific, we innovatively develop a new graph learning paradigm by leveraging an explicit theoretical connection between clustering indicator matrix and affinity graph, such that the affinity graph propagated from KKM enjoys the valuable block diagonal and sparse property. Then, by using this graph learning paradigm, base kernels can produce multiple candidate affinity graphs, which are stacked into a low-rank graph tensor for capturing the high-order affinity of all these graphs. After that, by averaging all the frontal slices of the tensor, a high-quality affinity graph is obtained. Extensive experiments have shown the superiority of KCGT compared with the state-of-the-art MKC methods.

Subspace clustering via stacked independent subspace analysis networks with sparse prior information

Sparse subspace clustering (SSC) method has gained considerable attention in recent decades owing to its advantages in the fields of clustering. In essence, SSC is to learn a sparse affinity matrix followed by striving for a low-dimensional representation of data. However, the SSC and its variants mainly focus on building high-quality affinity matrix while ignoring the importance of low-dimensional feature derived from the affinity matrix. Moreover, due to their intrinsic linearity of models, they cannot efficiently handle data with the nonlinear distribution. In this paper, we propose a stacked independent subspace analysis (ISA) with sparse prior information called stacked-ISASP to deal with these two issues. Powered by handling data with nonlinear structure, our method aims at seeking a low-dimensional feature from the image data. Concretely, the model can stack the modified independent subspace analysis networks by incorporating the prior subspace information from the original data. To validate the efficiency of the proposed method, we compare our proposed stacked-ISASP method with the state-of-the-art methods on real datasets. Experimental results show that our approach can not only learn a better low-dimensional structure from the data but also achieve better performance for the classification task.

Robust self-tuning spectral clustering

Clustering, as an effective data analysis technique, is widely used in industrial application and science research. In this paper, we have proposed a novel spectral clustering method to solve three problems of spectral clustering, i.e., cluster-initialization, cluster-specification, and noise-robustness. To do this, we first learn a high-quality affinity matrix, and then capture an inherent clustering pattern in real feature space and projected subspace, and finally extract connected components to conduct clustering. By comparing to five comparison methods in six real data sets, our proposed method has achieved a competitive clustering performance in terms of four evaluation metrics.

Selection of Antibody Fragments by Yeast Display.

The critical need for renewable, high-quality affinity reagents in biological research, as well as for diagnostic and therapeutic applications, has required the development of new platforms of discovery. Yeast display is one of the main methods of in vitro display technology with phage display. Yeast display has been chosen by numerous groups to refine both affinity and specificity of antibodies because it enables fine discrimination between mutant clones of similar affinity. In addition, the construction of display libraries of antibody fragments in yeast permits to sample the immune antibody repertoire more fully than using phage. This chapter gives an updated overview of the available systems of yeast display platforms and libraries, followed up by technical descriptions of selection methods of antibody fragments by yeast display.

Generating Recombinant Antibodies to Membrane Proteins through Phage Display.

One of the most important classes of proteins in terms of drug targets is cell surface membrane proteins, and yet it is a challenging set of proteins for generating high-quality affinity reagents. In this review, we focus on the use of phage libraries, which display antibody fragments, for generating recombinant antibodies to membrane proteins. Such affinity reagents generally have high specificity and affinity for their targets. They have been used for cell staining, for promoting protein crystallization to solve three-dimensional structures, for diagnostics, and for treating diseases as therapeutics. We cover publications on this topic from the past 10 years, with a focus on the various formats of membrane proteins for affinity selection and the diverse affinity selection strategies used. Lastly, we discuss the challenges faced in this field and provide possible directions for future efforts.

Directed Evolution of a Highly Specific FN3 Monobody to the SH3 Domain of Human Lyn Tyrosine Kinase.

Affinity reagents of high affinity and specificity are very useful for studying the subcellular locations and quantities of individual proteins. To generate high-quality affinity reagents for human Lyn tyrosine kinase, a phage display library of fibronectin type III (FN3) monobodies was affinity selected with a recombinant form of the Lyn SH3 domain. While a highly specific monobody, TA8, was initially isolated, we chose to improve its affinity through directed evolution. A secondary library of 1.2 × 109 variants was constructed and screened by affinity selection, yielding three variants, two of which have affinities of ~ 40 nM, a 130-fold increase over the original TA8 monobody. One of the variants, 2H7, displayed high specificity to the Lyn SH3 domain, as shown by ELISA and probing arrays of 150 SH3 domains. Furthermore, the 2H7 monobody was able to pull down endogenous Lyn from a lysate of Burkitt's lymphoma cells, thereby demonstrating its utility as an affinity reagent for detecting Lyn in a complex biological mixture.

Quantitative selection and parallel characterization of aptamers

Aptamers are promising affinity reagents that are potentially well suited for high-throughput discovery, as they are chemically synthesized and discovered via completely in vitro selection processes. Recent advancements in selection, sequencing, and the use of modified bases have improved aptamer quality, but the overall process of aptamer generation remains laborious and low-throughput. This is because binding characterization remains a critical bottleneck, wherein the affinity and specificity of each candidate aptamer are measured individually in a serial manner. To accelerate aptamer discovery, we devised the Quantitative Parallel Aptamer Selection System (QPASS), which integrates microfluidic selection and next-generation sequencing with in situ-synthesized aptamer arrays, enabling simultaneous measurement of affinity and specificity for thousands of candidate aptamers in parallel. After using QPASS to select aptamers for the human cancer biomarker angiopoietin-2 (Ang2), we in situ synthesized arrays of the selected sequences and obtained equilibrium dissociation constants (Kd) for every aptamer in parallel. We thereby identified over a dozen high-affinity Ang2 aptamers, with Kd as low as 20.5 ± 7.3 nM. The same arrays enabled us to quantify binding specificity for these aptamers in parallel by comparing relative binding of differentially labeled target and nontarget proteins, and by measuring their binding affinity directly in complex samples such as undiluted serum. Finally, we show that QPASS offers a compelling avenue for exploring structure-function relationships for large numbers of aptamers in parallel by coupling array-based affinity measurements with next-generation sequencing data to identify nucleotides and motifs within the aptamer that critically affect Ang2 binding.