Computational Biomarker Pipeline from Discovery to Clinical Implementation: Plasma Proteomic Biomarkers for Cardiac Transplantation

Recent technical advances in the field of quantitative proteomics have stimulated a large number of biomarker discovery studies of various diseases, providing avenues for new treatments and diagnostics. However, inherent challenges have limited the successful translation of candidate biomarkers into clinical use, thus highlighting the need for a robust analytical methodology to transition from biomarker discovery to clinical implementation. We have developed an end-to-end computational proteomic pipeline for biomarkers studies. At the discovery stage, the pipeline emphasizes different aspects of experimental design, appropriate statistical methodologies, and quality assessment of results. At the validation stage, the pipeline focuses on the migration of the results to a platform appropriate for external validation, and the development of a classifier score based on corroborated protein biomarkers. At the last stage towards clinical implementation, the main aims are to develop and validate an assay suitable for clinical deployment, and to calibrate the biomarker classifier using the developed assay. The proposed pipeline was applied to a biomarker study in cardiac transplantation aimed at developing a minimally invasive clinical test to monitor acute rejection. Starting with an untargeted screening of the human plasma proteome, five candidate biomarker proteins were identified. Rejection-regulated proteins reflect cellular and humoral immune responses, acute phase inflammatory pathways, and lipid metabolism biological processes. A multiplex multiple reaction monitoring mass-spectrometry (MRM-MS) assay was developed for the five candidate biomarkers and validated by enzyme-linked immune-sorbent (ELISA) and immunonephelometric assays (INA). A classifier score based on corroborated proteins demonstrated that the developed MRM-MS assay provides an appropriate methodology for an external validation, which is still in progress. Plasma proteomic biomarkers of acute cardiac rejection may offer a relevant post-transplant monitoring tool to effectively guide clinical care. The proposed computational pipeline is highly applicable to a wide range of biomarker proteomic studies.     

Discriminative motifs.

This paper takes a new view of motif discovery, addressing a common problem in existing motif finders. A motif is treated as a feature of the input promoter regions that leads to a good classifier between these promoters and a set of background promoters. This perspective allows us to adapt existing methods of feature selection, a well-studied topic in machine learning, to motif discovery. We develop a general algorithmic framework that can be specialized to work with a wide variety of motif models, including consensus models with degenerate symbols or mismatches, and composite motifs. A key feature of our algorithm is that it measures overrepresentation while maintaining information about the distribution of motif instances in individual promoters. The assessment of a motif's discriminative power is normalized against chance behaviour by a probabilistic analysis. We apply our framework to two popular motif models and are able to detect several known binding sites in sets of co-regulated genes in yeast.     

Computational identification of human long intergenic non-coding RNAs using a GA–SVM algorithm

Long intergenic non-coding RNAs (lincRNAs) are a new type of non-coding RNAs and are closely related with the occurrence and development of diseases. In previous studies, most lincRNAs have been identified through next-generation sequencing. Because lincRNAs exhibit tissue-specific expression, the reproducibility of lincRNA discovery in different studies is very poor. In this study, not including lincRNA expression, we used the sequence, structural and protein-coding potential features as potential features to construct a classifier that can be used to distinguish lincRNAs from non-lincRNAs. The GA–SVM algorithm was performed to extract the optimized feature subset. Compared with several feature subsets, the five-fold cross validation results showed that this optimized feature subset exhibited the best performance for the identification of human lincRNAs. Moreover, the LincRNA Classifier based on Selected Features (linc-SF) was constructed by support vector machine (SVM) based on the optimized feature subset. The performance of this classifier was further evaluated by predicting lincRNAs from two independent lincRNA sets. Because the recognition rates for the two lincRNA sets were 100% and 99.8%, the linc-SF was found to be effective for the prediction of human lincRNAs.     

SLiMPrints: conservation-based discovery of functional motif fingerprints in intrinsically disordered protein regions

Large portions of higher eukaryotic proteomes are intrinsically disordered, and abundant evidence suggests that these unstructured regions of proteins are rich in regulatory interaction interfaces. A major class of disordered interaction interfaces are the compact and degenerate modules known as short linear motifs (SLiMs). As a result of the difficulties associated with the experimental identification and validation of SLiMs, our understanding of these modules is limited, advocating the use of computational methods to focus experimental discovery. This article evaluates the use of evolutionary conservation as a discriminatory technique for motif discovery. A statistical framework is introduced to assess the significance of relatively conserved residues, quantifying the likelihood a residue will have a particular level of conservation given the conservation of the surrounding residues. The framework is expanded to assess the significance of groupings of conserved residues, a metric that forms the basis of SLiMPrints (short linear motif fingerprints), a de novo motif discovery tool. SLiMPrints identifies relatively overconstrained proximal groupings of residues within intrinsically disordered regions, indicative of putatively functional motifs. Finally, the human proteome is analysed to create a set of highly conserved putative motif instances, including a novel site on translation initiation factor eIF2A that may regulate translation through binding of eIF4E.     

Statistical data processing in clinical proteomics

This review discusses data analysis strategies for the discovery of biomarkers in clinical proteomics. Proteomics studies produce large amounts of data, characterized by few samples of which many variables are measured. A wealth of classification methods exists for extracting information from the data. Feature selection plays an important role in reducing the dimensionality of the data prior to classification and in discovering biomarker leads. The question which classification strategy works best is yet unanswered. Validation is a crucial step for biomarker leads towards clinical use. Here we only discuss statistical validation, recognizing that biological and clinical validation is of utmost importance. First, there is the need for validated model selection to develop a generalized classifier that predicts new samples correctly. A cross-validation loop that is wrapped around the model development procedure assesses the performance using unseen data. The significance of the model should be tested; we use permutations of the data for comparison with uninformative data. This procedure also tests the correctness of the performance validation. Preferably, a new set of samples is measured to test the classifier and rule out results specific for a machine, analyst, laboratory or the first set of samples. This is not yet standard practice. We present a modular framework that combines feature selection, classification, biomarker discovery and statistical validation; these data analysis aspects are all discussed in this review. The feature selection, classification and biomarker discovery modules can be incorporated or omitted to the preference of the researcher. The validation modules, however, should not be optional. In each module, the researcher can select from a wide range of methods, since there is not one unique way that leads to the correct model and proper validation. We discuss many possibilities for feature selection, classification and biomarker discovery. For validation we advice a combination of cross-validation and permutation testing, a validation strategy supported in the literature.     

miRDetect: A combinatorial approach for automated detection of novel miRNA precursors from plant EST data using homology and Random Forest classification

Identification of microRNAs from plants is a crucial step for understanding the mechanisms of pathways and regulation of genes. A number of tools have been developed for the detection of microRNAs from small RNA-seq data. However, there is a lack of pipeline for detection of miRNA from EST dataset even when a huge resource is publicly available and the method is known. Here we present miRDetect, a python implementation to detect novel miRNA precursors from plant EST data using homology and machine learning approach. 10-fold cross validation was applied to choose best classifier based on ROC, accuracy, MCC and F1-scores using 112 features. miRDetect achieved a classification accuracy of 93.35% on a Random Forest classifier and outperformed other precursor detection tools in terms of performance. The miRDetect pipeline aids in identifying novel plant precursors using a mixed approach and will be helpful to researchers with less informatics background.     

A generic motif discovery algorithm for sequential data

MOTIVATION: Motif discovery in sequential data is a problem of great interest and with many applications. However, previous methods have been unable to combine exhaustive search with complex motif representations and are each typically only applicable to a certain class of problems. RESULTS: Here we present a generic motif discovery algorithm (Gemoda) for sequential data. Gemoda can be applied to any dataset with a sequential character, including both categorical and real-valued data. As we show, Gemoda deterministically discovers motifs that are maximal in composition and length. As well, the algorithm allows any choice of similarity metric for finding motifs. Finally, Gemoda's output motifs are representation-agnostic: they can be represented using regular expressions, position weight matrices or any number of other models for any type of sequential data. We demonstrate a number of applications of the algorithm, including the discovery of motifs in amino acids sequences, a new solution to the (l,d)-motif problem in DNA sequences and the discovery of conserved protein substructures. AVAILABILITY: Gemoda is freely available at http://web.mit.edu/bamel/gemoda     

Creating robust, reliable, clinically relevant classifiers from spectroscopic data

I describe in detail the intimately connected feature extraction and classifier development stages of the data-driven Statistical Classification Strategy (SCS) and compare them with current practice used in MR spectroscopy. We initially created the SCS for the analysis of MR and IR spectra of biofluids and tissues, and subsequently extended it to analyze biomedical data in general. I focus on explaining how to extract discriminatory spectral features and create robust classifiers that can reliably discriminate diseases and disease states. I discuss our approach to identifying features that retain spectral identity and provisionally relate these features, averaged subregions of the spectra, to specific chemical entities (“metabolites”). Particular emphasis is placed on describing the steps required to help create classifiers whose accuracy doesn’t deteriorate significantly when presented with new, unknown samples. A simple but powerful extension of the discovered features to detect metabolite-metabolite (feature-feature) interactions is also sketched. I contrast the advantages and disadvantages of using either spectral signatures or explicit metabolite concentrations derived from the spectra as sets of discriminatory features. At present, no clear-cut preference is obvious and more objective comparisons will be needed. Finally, I argue that clinical requirements and exigencies strongly suggest adopting a two-phase approach to diagnosis/prognosis. In the first phase the emphasis ought to be on providing as accurate a diagnosis as possible, without any attempt to identify “biomarkers.” That should be the goal of the second, research phase, with a view of providing prognosis on disease progression.     

AMASS: algorithm for MSI analysis by semi-supervised segmentation

Mass Spectrometric Imaging (MSI) is a molecular imaging technique that allows the generation of 2D ion density maps for a large complement of the active molecules present in cells and sectioned tissues. Automatic segmentation of such maps according to patterns of co-expression of individual molecules can be used for discovery of novel molecular signatures (molecules that are specifically expressed in particular spatial regions). However, current segmentation techniques are biased toward the discovery of higher abundance molecules and large segments; they allow limited opportunity for user interaction, and validation is usually performed by similarity to known anatomical features. We describe here a novel method, AMASS (Algorithm for MSI Analysis by Semi-supervised Segmentation). AMASS relies on the discriminating power of a molecular signal instead of its intensity as a key feature, uses an internal consistency measure for validation, and allows significant user interaction and supervision as options. An automated segmentation of entire leech embryo data images resulted in segmentation domains congruent with many known organs, including heart, CNS ganglia, nephridia, nephridiopores, and lateral and ventral regions, each with a distinct molecular signature. Likewise, segmentation of a rat brain MSI slice data set yielded known brain features and provided interesting examples of co-expression between distinct brain regions. AMASS represents a new approach for the discovery of peptide masses with distinct spatial features of expression. Software source code and installation and usage guide are available at http://bix.ucsd.edu/AMASS/ .     

Feature Selection and the Class Imbalance Problem in Predicting Protein Function from Sequence

When the standard approach to predict protein function by sequence homology fails, other alternative methods can be used that require only the amino acid sequence for predicting function. One such approach uses machine learning to predict protein function directly from amino acid sequence features. However, there are two issues to consider before successful functional prediction can take place: identifying discriminatory features, and overcoming the challenge of a large imbalance in the training data. We show that by applying feature subset selection followed by undersampling of the majority class, significantly better support vector machine (SVM) classifiers are generated compared with standard machine learning approaches. As well as revealing that the features selected could have the potential to advance our understanding of the relationship between sequence and function, we also show that undersampling to produce fully balanced data significantly improves performance. The best discriminating ability is achieved using SVMs together with feature selection and full undersampling; this approach strongly outperforms other competitive learning algorithms. We conclude that this combined approach can generate powerful machine learning classifiers for predicting protein function directly from sequence.     

Enhanced Regulatory Sequence Prediction Using Gapped k-mer Features

Oligomers of length k, or k-mers, are convenient and widely used features for modeling the properties and functions of DNA and protein sequences. However, k-mers suffer from the inherent limitation that if the parameter k is increased to resolve longer features, the probability of observing any specific k-mer becomes very small, and k-mer counts approach a binary variable, with most k-mers absent and a few present once. Thus, any statistical learning approach using k-mers as features becomes susceptible to noisy training set k-mer frequencies once k becomes large. To address this problem, we introduce alternative feature sets using gapped k-mers, a new classifier, gkm-SVM, and a general method for robust estimation of k-mer frequencies. To make the method applicable to large-scale genome wide applications, we develop an efficient tree data structure for computing the kernel matrix. We show that compared to our original kmer-SVM and alternative approaches, our gkm-SVM predicts functional genomic regulatory elements and tissue specific enhancers with significantly improved accuracy, increasing the precision by up to a factor of two. We then show that gkm-SVM consistently outperforms kmer-SVM on human ENCODE ChIP-seq datasets, and further demonstrate the general utility of our method using a Naïve-Bayes classifier. Although developed for regulatory sequence analysis, these methods can be applied to any sequence classification problem.     

Support Vector Machine-based classification of protein folds using the structural properties of amino acid residues and amino acid residue pairs

MOTIVATION: Fold recognition is a key step in the protein structure discovery process, especially when traditional sequence comparison methods fail to yield convincing structural homologies. Although many methods have been developed for protein fold recognition, their accuracies remain low. This can be attributed to insufficient exploitation of fold discriminatory features. RESULTS: We have developed a new method for protein fold recognition using structural information of amino acid residues and amino acid residue pairs. Since protein fold recognition can be treated as a protein fold classification problem, we have developed a Support Vector Machine (SVM) based classifier approach that uses secondary structural state and solvent accessibility state frequencies of amino acids and amino acid pairs as feature vectors. Among the individual properties examined secondary structural state frequencies of amino acids gave an overall accuracy of 65.2% for fold discrimination, which is better than the accuracy by any method reported so far in the literature. Combination of secondary structural state frequencies with solvent accessibility state frequencies of amino acids and amino acid pairs further improved the fold discrimination accuracy to more than 70%, which is approximately 8% higher than the best available method. In this study we have also tested, for the first time, an all-together multi-class method known as Crammer and Singer method for protein fold classification. Our studies reveal that the three multi-class classification methods, namely one versus all, one versus one and Crammer and Singer method, yield similar predictions. AVAILABILITY: Dataset and stand-alone program are available upon request.     

Exploiting a reduced set of weighted average features to improve prediction of DNA-binding residues from 3D structures

Predicting DNA-binding residues from a protein three-dimensional structure is a key task of computational structural proteomics. In the present study, based on machine learning technology, we aim to explore a reduced set of weighted average features for improving prediction of DNA-binding residues on protein surfaces. Via constructing the spatial environment around a DNA-binding residue, a novel weighting factor is first proposed to quantify the distance-dependent contribution of each neighboring residue in determining the location of a binding residue. Then, a weighted average scheme is introduced to represent the surface patch of the considering residue. Finally, the classifier is trained on the reduced set of these weighted average features, consisting of evolutionary profile, interface propensity, betweenness centrality and solvent surface area of side chain. Experimental results on 5-fold cross validation and independent tests indicate that the new feature set are effective to describe DNA-binding residues and our approach has significantly better performance than two previous methods. Furthermore, a brief case study suggests that the weighted average features are powerful for identifying DNA-binding residues and are promising for further study of protein structure-function relationship. The source code and datasets are available upon request.     

Coherent pipeline for biomarker discovery using mass spectrometry and bioinformatics

Background  

Robust biomarkers are needed to improve microbial identification and diagnostics. Proteomics methods based on mass spectrometry can be used for the discovery of novel biomarkers through their high sensitivity and specificity. However, there has been a lack of a coherent pipeline connecting biomarker discovery with established approaches for evaluation and validation. We propose such a pipeline that uses in silico methods for refined biomarker discovery and confirmation.     

Prediction of protein‐glucose binding sites using support vector machines

Glucose is a simple sugar that plays an essential role in many basic metabolic and signaling pathways. Many proteins have binding sites that are highly specific to glucose. The exponential increase of genomic data has revealed the identity of many proteins that seem to be central to biological processes, but whose exact functions are unknown. Many of these proteins seem to be associated with disease processes. Being able to predict glucose‐specific binding sites in these proteins will greatly enhance our ability to annotate protein function and may significantly contribute to drug design. We hereby present the first glucose‐binding site classifier algorithm. We consider the sugar‐binding pocket as a spherical spatio‐chemical environment and represent it as a vector of geometric and chemical features. We then perform Random Forests feature selection to identify key features and analyze them using support vector machines classification. Our work shows that glucose binding sites can be modeled effectively using a limited number of basic chemical and residue features. Using a leave‐one‐out cross‐validation method, our classifier achieves a 8.11% error, a 89.66% sensitivity and a 93.33% specificity over our dataset. From a biochemical perspective, our results support the relevance of ordered water molecules and ions in determining glucose specificity. They also reveal the importance of carboxylate residues in glucose binding and the high concentration of negatively charged atoms in direct contact with the bound glucose molecule. Proteins 2009. © 2009 Wiley‐Liss, Inc.