Ching-Tai Chen

Rationalization and Design of the Complementarity Determining Region Sequences in an Antibody-Antigen Recognition Interface

Protein-protein interactions are critical determinants in biological systems. Engineered proteins binding to specific areas on protein surfaces could lead to therapeutics or diagnostics for treating diseases in humans. But designing epitope-specific protein-protein interactions with computational atomistic interaction free energy remains a difficult challenge. Here we show that, with the antibody-VEGF (vascular endothelial growth factor) interaction as a model system, the experimentally observed amino acid preferences in the antibody-antigen interface can be rationalized with 3-dimensional distributions of interacting atoms derived from the database of protein structures. Machine learning models established on the rationalization can be generalized to design amino acid preferences in antibody-antigen interfaces, for which the experimental validations are tractable with current high throughput synthetic antibody display technologies. Leave-one-out cross validation on the benchmark system yielded the accuracy, precision, recall (sensitivity) and specificity of the overall binary predictions to be 0.69, 0.45, 0.63, and 0.71 respectively, and the overall Matthews correlation coefficient of the 20 amino acid types in the 24 interface CDR positions was 0.312. The structure-based computational antibody design methodology was further tested with other antibodies binding to VEGF. The results indicate that the methodology could provide alternatives to the current antibody technologies based on animal immune systems in engineering therapeutic and diagnostic antibodies against predetermined antigen epitopes.

Protein-Protein Interaction Site Predictions with Three-Dimensional Probability Distributions of Interacting Atoms on Protein Surfaces

Protein-protein interactions are key to many biological processes. Computational methodologies devised to predict protein-protein interaction (PPI) sites on protein surfaces are important tools in providing insights into the biological functions of proteins and in developing therapeutics targeting the protein-protein interaction sites. One of the general features of PPI sites is that the core regions from the two interacting protein surfaces are complementary to each other, similar to the interior of proteins in packing density and in the physicochemical nature of the amino acid composition. In this work, we simulated the physicochemical complementarities by constructing three-dimensional probability density maps of non-covalent interacting atoms on the protein surfaces. The interacting probabilities were derived from the interior of known structures. Machine learning algorithms were applied to learn the characteristic patterns of the probability density maps specific to the PPI sites. The trained predictors for PPI sites were cross-validated with the training cases (consisting of 432 proteins) and were tested on an independent dataset (consisting of 142 proteins). The residue-based Matthews correlation coefficient for the independent test set was 0.423; the accuracy, precision, sensitivity, specificity were 0.753, 0.519, 0.677, and 0.779 respectively. The benchmark results indicate that the optimized machine learning models are among the best predictors in identifying PPI sites on protein surfaces. In particular, the PPI site prediction accuracy increases with increasing size of the PPI site and with increasing hydrophobicity in amino acid composition of the PPI interface; the core interface regions are more likely to be recognized with high prediction confidence. The results indicate that the physicochemical complementarity patterns on protein surfaces are important determinants in PPIs, and a substantial portion of the PPI sites can be predicted correctly with the physicochemical complementarity features based on the non-covalent interaction data derived from protein interiors.

Protease substrate site predictors derived from machine learning on multilevel substrate phage display data

MOTIVATION Regulatory proteases modulate proteomic dynamics with a spectrum of specificities against substrate proteins. Predictions of the substrate sites in a proteome for the proteases would facilitate understanding the biological functions of the proteases. High-throughput experiments could generate suitable datasets for machine learning to grasp complex relationships between the substrate sequences and the enzymatic specificities. But the capability in predicting protease substrate sites by integrating the machine learning algorithms with the experimental methodology has yet to be demonstrated. RESULTS Factor Xa, a key regulatory protease in the blood coagulation system, was used as model system, for which effective substrate site predictors were developed and benchmarked. The predictors were derived from bootstrap aggregation (machine learning) algorithms trained with data obtained from multilevel substrate phage display experiments. The experimental sampling and computational learning on substrate specificities can be generalized to proteases for which the active forms are available for the in vitro experiments. AVAILABILITY http://asqa.iis.sinica.edu.tw/fXaWeb/

MAGIC-web: a platform for untargeted and targeted N-linked glycoprotein identification.

MAGIC-web is the first web server, to the best of our knowledge, that performs both untargeted and targeted analyses of mass spectrometry-based glycoproteomics data for site-specific N-linked glycoprotein identification. The first two modules, MAGIC and MAGIC+, are designed for untargeted and targeted analysis, respectively. MAGIC is implemented with our previously proposed novel Y1-ion pattern matching method, which adequately detects Y1- and Y0-ion without prior information of proteins and glycans, and then generates in silico MS2 spectra that serve as input to a database search engine (e.g. Mascot) to search against a large-scale protein sequence database. On top of that, the newly implemented MAGIC+ allows users to determine glycopeptide sequences using their own protein sequence file. The third module, Reports Integrator, provides the service of combining protein identification results from Mascot and glycan-related information from MAGIC-web to generate a complete site-specific protein-glycan summary report. The last module, Glycan Search, is designed for the users who are interested in finding possible glycan structures with specific numbers and types of monosaccharides. The results from MAGIC, MAGIC+ and Reports Integrator can be downloaded via provided links whereas the annotated spectra and glycan structures can be visualized in the browser. MAGIC-web is accessible from http://ms.iis.sinica.edu.tw/MAGIC-web/index.html.

Various high-order modes in vertical-cavity surface-emitting lasers with equilateral triangular lateral confinement

Large-aperture vertical-cavity surface-emitting lasers with an equilateral triangular lateral confinement are fabricated to investigate the formation of high-order resonant modes. The experimental lasing patterns are composed of the superscar mode, honeycomb eigenstate, and chaotic mode. Experimental results confirm the theoretical predictions that tiny symmetry breaking can cause the high-order modes to reveal miscellaneous states of integrable and chaotic systems.

HYPLOSP: A KNOWLEDGE-BASED APPROACH TO PROTEIN LOCAL STRUCTURE PREDICTION

Local structure prediction can facilitate ab initio structure prediction, protein threading, and remote homology detection. However, the accuracy of existing methods is limited. In this paper, we propose a knowledge-based prediction method that assigns a measure called the local match rate to each position of an amino acid sequence to estimate the confidence of our method. Empirically, the accuracy of the method correlates positively with the local match rate; therefore, we employ it to predict the local structures of positions with a high local match rate. For positions with a low local match rate, we propose a neural network prediction method. To better utilize the knowledge-based and neural network methods, we design a hybrid prediction method, HYPLOSP (HYbrid method to Protein LOcal Structure Prediction) that combines both methods. To evaluate the performance of the proposed methods, we first perform cross-validation experiments by applying our knowledge-based method, a neural network method, and HYPLOSP to a large dataset of 3,925 protein chains. We test our methods extensively on three different structural alphabets and evaluate their performance by two widely used criteria, Maximum Deviation of backbone torsion Angle (MDA) and Q(N), which is similar to Q(3) in secondary structure prediction. We then compare HYPLOSP with three previous studies using a dataset of 56 new protein chains. HYPLOSP shows promising results in terms of MDA and Q(N) accuracy and demonstrates its alphabet-independent capability.

iMet-Q: A User-Friendly Tool for Label-Free Metabolomics Quantitation Using Dynamic Peak-Width Determination

Efficient and accurate quantitation of metabolites from LC-MS data has become an important topic. Here we present an automated tool, called iMet-Q (intelligent Metabolomic Quantitation), for label-free metabolomics quantitation from high-throughput MS1 data. By performing peak detection and peak alignment, iMet-Q provides a summary of quantitation results and reports ion abundance at both replicate level and sample level. Furthermore, it gives the charge states and isotope ratios of detected metabolite peaks to facilitate metabolite identification. An in-house standard mixture and a public Arabidopsis metabolome data set were analyzed by iMet-Q. Three public quantitation tools, including XCMS, MetAlign, and MZmine 2, were used for performance comparison. From the mixture data set, seven standard metabolites were detected by the four quantitation tools, for which iMet-Q had a smaller quantitation error of 12% in both profile and centroid data sets. Our tool also correctly determined the charge states of seven standard metabolites. By searching the mass values for those standard metabolites against Human Metabolome Database, we obtained a total of 183 metabolite candidates. With the isotope ratios calculated by iMet-Q, 49% (89 out of 183) metabolite candidates were filtered out. From the public Arabidopsis data set reported with two internal standards and 167 elucidated metabolites, iMet-Q detected all of the peaks corresponding to the internal standards and 167 metabolites. Meanwhile, our tool had small abundance variation (≤0.19) when quantifying the two internal standards and had higher abundance correlation (≥0.92) when quantifying the 167 metabolites. iMet-Q provides user-friendly interfaces and is publicly available for download at http://ms.iis.sinica.edu.tw/comics/Software_iMet-Q.html.

A Knowledge-Based Approach to Protein Local Structure Prediction.

Local structure prediction can facilitate ab initio structure prediction, protein threading, and remote homology detection. However, previous approaches to local structure prediction suffer from poor accuracy. In this paper, we propose a knowledge-based prediction method that assigns a measure called the local match rate to each position of an amino acid sequence to estimate the confidence of our approach. To remedy prediction results with low local match rates, we use a neural network prediction method. Then, we have a hybrid prediction method, HYPLOSP (HYbrid method to Protein LOcal Structure Prediction) that combines our knowledge-based method with a neural network method. We test the method on two different structural alphabets and evaluate it by QN, which is similar to Q3 in secondary structure prediction. The experimental results show that our method yields a significant improvement over previous studies.

Design of an agent-based framework for processes collaboration in electronic marketplace

Specialised business-to-business Internet electronic markets deliver substantial value to companies including greater liquidity, better pricing, and faster transactions. The electronic marketplace sits on the intersection between supply chain and demand chain and makes it the business collaborative platform where resource allocation optimising and efficiency improving occurs simultaneously. The electronic marketplaces can help companies reduce cost structures by accessing stable venues with market pricing to fulfil short-term commodity needs. However, the variances in business processes and data format between trading partners cause overhead cost and time consumption. This paper proposes an agent-based framework to integrate processes between businesses and the electronic marketplace. This framework is composed of workflow systems, collaborative agents, Common Business Language, and electronic marketplace service platform. The business vocabularies and inter-partner processes used to communicate between trading partners are defined as an XML-based format. The result of this framework provides three functions including: (1) process executing; (2) message transforming; and (3) business transactions tracking and management. Trading partners can integrate inter-organisational processes without dramatically rewriting their legacy system.

UniLoc: A universal protein localization site predictor for eukaryotes and prokaryotes

There is a growing gap between protein subcellular localization (PSL) data and protein sequence data, raising the need for computation methods to rapidly determine subcellular localizations for uncharacterized proteins. Currently, the most efficient computation method involves finding sequence-similar proteins (hereafter referred to as similar proteins) in the annotated database and transferring their annotations to the target protein. When a sequence-similarity search fails to find similar proteins, many PSL predictors adopt machine learning methods for the prediction of localization sites. We proposed a universal protein localization site predictor - UniLoc - to take advantage of implicit similarity among proteins through sequence analysis alone. The notion of related protein words is introduced to explore the localization site assignment of uncharacterized proteins. UniLoc is found to identify useful template proteins and produce reliable predictions when similar proteins were not available.