Classification of nucleotide sequences over their frequency dictionaries reveals a relation between the structure of sequences and taxonomy of their bearers

Classification of 16S RNA sequences over their frequency dictionaries, both real ones, and transformed ones was studied. Two entities were considered to be close each other from the point of view of their structure, if their frequency dictionaries were close, in Eucledian metric. A transformation procedure of a frequency dictionary has been implemented that reveals the peculiarities of information structure of a nucleotide sequence. A comparative study of two classification developed over the real frequency dictionary vs. that one developed over the transformed frequency dictionary was carried out. The strong correlation is revealed between the classification and the taxonomy of 16S RNA bearer. For the classes isolated, the information valuable words were identified. These words are the main factors of a difference between the classes. The frequency dictionaries containing the words of the length 3 exhibit the best correlation between a class and a genus. A genus, as a rule, is included into the same class, and the exclusion are sporadic. A development of hierarchy classification over the transformed frequency dictionaries separated one or two taxonomy groups, as each stage of classification. The unexpectedly frequent, or contrary, unexpectedly rare occurred of words (of the length 3) in entities under consideration make the structure difference between the classes of the nucleotide sequences.     

A compression-based approach for coding sequences identification. I. Application to prokaryotic genomes.

Most of the gene prediction algorithms for prokaryotes are based on Hidden Markov Models or similar machine-learning approaches, which imply the optimization of a high number of parameters. The present paper presents a novel method for the classification of coding and non-coding regions in prokaryotic genomes, based on a suitably defined compression index of a DNA sequence. The main features of this new method are the non-parametric logic and the costruction of a dictionary of words extracted from the sequences. These dictionaries can be very useful to perform further analyses on the genomic sequences themselves. The proposed approach has been applied on some prokaryotic complete genomes, obtaining optimal scores of correctly recognized coding and non-coding regions. Several false-positive and false-negative cases have been investigated in detail, which have revealed that this approach can fail in the presence of highly structured coding regions (e.g., genes coding for modular proteins) or quasi-random non-coding regions (e.g., regions hosting non-functional fragments of copies of functional genes; regions hosting promoters or other protein-binding sequences). We perform an overall comparison with other gene-finder software, since at this step we are not interested in building another gene-finder system, but only in exploring the possibility of the suggested approach.     

Automated hierarchical classification of protein domain subfamilies based on functionally-divergent residue signatures

ABSTRACT: BACKGROUND: The NCBI Conserved Domain Database (CDD) consists of a collection of multiple sequence alignments of protein domains that are at various stages of being manually curated into evolutionary hierarchies based on conserved and divergent sequence and structural features. These domain models are annotated to provide insights into the relationships between sequence, structure and function via web-based BLAST searches. RESULTS: Here we automate the generation of conserved domain (CD) hierarchies using a combination of heuristic and Markov chain Monte Carlo sampling procedures and starting from a (typically very large) multiple sequence alignment. This procedure relies on statistical criteria to define each hierarchy based on the conserved and divergent sequence patterns associated with protein functional-specialization. At the same time this facilitates the sequence and structural annotation of residues that are functionally important. These statistical criteria also provide a means to objectively assess the quality of CD hierarchies, a non-trivial task considering that the protein subgroups are often very distantly related--a situation in which standard phylogenetic methods can be unreliable. Our aim here is to automatically generate (typically sub-optimal) hierarchies that, based on statistical criteria and visual comparisons, are comparable to manually curated hierarchies; this serves as the first step toward the ultimate goal of obtaining optimal hierarchical classifications. A plot of runtimes for the most time-intensive (non-parallelizable) part of the algorithm indicates a nearly linear time complexity so that, even for the extremely large Rossmann fold protein class, results are obtained in about a day. CONCLUSIONS: This approach automates the rapid creation of protein domain hierarchies and thus will eliminate one of the most time consuming aspects of conserved domain database curation. At the same time, it also facilitates protein domain annotation by identifying those pattern residues that most distinguish each protein domain subgroup from other related subgroups.     

Mass type-specific sparse representation for mass classification in computer-aided detection on mammograms

Background

Breast cancer is the leading cause of both incidence and mortality in women population. For this reason, much research effort has been devoted to develop Computer-Aided Detection (CAD) systems for early detection of the breast cancers on mammograms. In this paper, we propose a new and novel dictionary configuration underpinning sparse representation based classification (SRC). The key idea of the proposed algorithm is to improve the sparsity in terms of mass margins for the purpose of improving classification performance in CAD systems.

Methods

The aim of the proposed SRC framework is to construct separate dictionaries according to the types of mass margins. The underlying idea behind our method is that the separated dictionaries can enhance the sparsity of mass class (true-positive), leading to an improved performance for differentiating mammographic masses from normal tissues (false-positive). When a mass sample is given for classification, the sparse solutions based on corresponding dictionaries are separately solved and combined at score level. Experiments have been performed on both database (DB) named as Digital Database for Screening Mammography (DDSM) and clinical Full Field Digital Mammogram (FFDM) DBs. In our experiments, sparsity concentration in the true class (SCTC) and area under the Receiver operating characteristic (ROC) curve (AUC) were measured for the comparison between the proposed method and a conventional single dictionary based approach. In addition, a support vector machine (SVM) was used for comparing our method with state-of-the-arts classifier extensively used for mass classification.

Results

Comparing with the conventional single dictionary configuration, the proposed approach is able to improve SCTC of up to 13.9% and 23.6% on DDSM and FFDM DBs, respectively. Moreover, the proposed method is able to improve AUC with 8.2% and 22.1% on DDSM and FFDM DBs, respectively. Comparing to SVM classifier, the proposed method improves AUC with 2.9% and 11.6% on DDSM and FFDM DBs, respectively.

Conclusions

The proposed dictionary configuration is found to well improve the sparsity of dictionaries, resulting in an enhanced classification performance. Moreover, the results show that the proposed method is better than conventional SVM classifier for classifying breast masses subject to various margins from normal tissues.

     

Identification of biological relationships from text documents using efficient computational methods

The biological literature databases continue to grow rapidly with vital information that is important for conducting sound biomedical research and development. The current practices of manually searching for information and extracting pertinent knowledge are tedious, time-consuming tasks even for motivated biological researchers. Accurate and computationally efficient approaches in discovering relationships between biological objects from text documents are important for biologists to develop biological models. The term "object" refers to any biological entity such as a protein, gene, cell cycle, etc. and relationship refers to any dynamic action one object has on another, e.g. protein inhibiting another protein or one object belonging to another object such as, the cells composing an organ. This paper presents a novel approach to extract relationships between multiple biological objects that are present in a text document. The approach involves object identification, reference resolution, ontology and synonym discovery, and extracting object-object relationships. Hidden Markov Models (HMMs), dictionaries, and N-Gram models are used to set the framework to tackle the complex task of extracting object-object relationships. Experiments were carried out using a corpus of one thousand Medline abstracts. Intermediate results were obtained for the object identification process, synonym discovery, and finally the relationship extraction. For the thousand abstracts, 53 relationships were extracted of which 43 were correct, giving a specificity of 81 percent. These results are promising for multi-object identification and relationship finding from biological documents.     

Extraction of Pharmacokinetic Evidence of Drug–Drug Interactions from the Literature

Drug-drug interaction (DDI) is a major cause of morbidity and mortality and a subject of intense scientific interest. Biomedical literature mining can aid DDI research by extracting evidence for large numbers of potential interactions from published literature and clinical databases. Though DDI is investigated in domains ranging in scale from intracellular biochemistry to human populations, literature mining has not been used to extract specific types of experimental evidence, which are reported differently for distinct experimental goals. We focus on pharmacokinetic evidence for DDI, essential for identifying causal mechanisms of putative interactions and as input for further pharmacological and pharmacoepidemiology investigations. We used manually curated corpora of PubMed abstracts and annotated sentences to evaluate the efficacy of literature mining on two tasks: first, identifying PubMed abstracts containing pharmacokinetic evidence of DDIs; second, extracting sentences containing such evidence from abstracts. We implemented a text mining pipeline and evaluated it using several linear classifiers and a variety of feature transforms. The most important textual features in the abstract and sentence classification tasks were analyzed. We also investigated the performance benefits of using features derived from PubMed metadata fields, various publicly available named entity recognizers, and pharmacokinetic dictionaries. Several classifiers performed very well in distinguishing relevant and irrelevant abstracts (reaching F1≈0.93, MCC≈0.74, iAUC≈0.99) and sentences (F1≈0.76, MCC≈0.65, iAUC≈0.83). We found that word bigram features were important for achieving optimal classifier performance and that features derived from Medical Subject Headings (MeSH) terms significantly improved abstract classification. We also found that some drug-related named entity recognition tools and dictionaries led to slight but significant improvements, especially in classification of evidence sentences. Based on our thorough analysis of classifiers and feature transforms and the high classification performance achieved, we demonstrate that literature mining can aid DDI discovery by supporting automatic extraction of specific types of experimental evidence.     

FuzzyART neural network for protein classification

One of the major research directions in bioinformatics is that of predicting the protein superfamily in large databases and classifying a given set of protein domains into superfamilies. The classification reflects the structural, evolutionary and functional relatedness. These relationships are embodied in hierarchical classification such as Structural Classification of Protein (SCOP), which is manually curated. Such classification is essential for the structural and functional analysis of proteins. Yet, a large number of proteins remain unclassified. We have proposed an unsupervised machine-learning FuzzyART neural network algorithm to classify a given set of proteins into SCOP superfamilies. The proposed method is fast learning and uses an atypical non-linear pattern recognition technique. In this approach, we have constructed a similarity matrix from p-values of BLAST all-against-all, trained the network with FuzzyART unsupervised learning algorithm using the similarity matrix as input vectors and finally the trained network offers SCOP superfamily level classification. In this experiment, we have evaluated the performance of our method with existing techniques on six different datasets. We have shown that the trained network is able to classify a given similarity matrix of a set of sequences into SCOP superfamilies at high classification accuracy.     

Sequence similarity network reveals common ancestry of multidomain proteins

We address the problem of homology identification in complex multidomain families with varied domain architectures. The challenge is to distinguish sequence pairs that share common ancestry from pairs that share an inserted domain but are otherwise unrelated. This distinction is essential for accuracy in gene annotation, function prediction, and comparative genomics. There are two major obstacles to multidomain homology identification: lack of a formal definition and lack of curated benchmarks for evaluating the performance of new methods. We offer preliminary solutions to both problems: 1) an extension of the traditional model of homology to include domain insertions; and 2) a manually curated benchmark of well-studied families in mouse and human. We further present Neighborhood Correlation, a novel method that exploits the local structure of the sequence similarity network to identify homologs with great accuracy based on the observation that gene duplication and domain shuffling leave distinct patterns in the sequence similarity network. In a rigorous, empirical comparison using our curated data, Neighborhood Correlation outperforms sequence similarity, alignment length, and domain architecture comparison. Neighborhood Correlation is well suited for automated, genome-scale analyses. It is easy to compute, does not require explicit knowledge of domain architecture, and classifies both single and multidomain homologs with high accuracy. Homolog predictions obtained with our method, as well as our manually curated benchmark and a web-based visualization tool for exploratory analysis of the network neighborhood structure, are available at http://www.neighborhoodcorrelation.org. Our work represents a departure from the prevailing view that the concept of homology cannot be applied to genes that have undergone domain shuffling. In contrast to current approaches that either focus on the homology of individual domains or consider only families with identical domain architectures, we show that homology can be rationally defined for multidomain families with diverse architectures by considering the genomic context of the genes that encode them. Our study demonstrates the utility of mining network structure for evolutionary information, suggesting this is a fertile approach for investigating evolutionary processes in the post-genomic era.     

A comparison of methods for classifying clinical samples based on proteomics data: a case study for statistical and machine learning approaches

The discovery of protein variation is an important strategy in disease diagnosis within the biological sciences. The current benchmark for elucidating information from multiple biological variables is the so called “omics” disciplines of the biological sciences. Such variability is uncovered by implementation of multivariable data mining techniques which come under two primary categories, machine learning strategies and statistical based approaches. Typically proteomic studies can produce hundreds or thousands of variables, p, per observation, n, depending on the analytical platform or method employed to generate the data. Many classification methods are limited by an n≪p constraint, and as such, require pre-treatment to reduce the dimensionality prior to classification. Recently machine learning techniques have gained popularity in the field for their ability to successfully classify unknown samples. One limitation of such methods is the lack of a functional model allowing meaningful interpretation of results in terms of the features used for classification. This is a problem that might be solved using a statistical model-based approach where not only is the importance of the individual protein explicit, they are combined into a readily interpretable classification rule without relying on a black box approach. Here we incorporate statistical dimension reduction techniques Partial Least Squares (PLS) and Principal Components Analysis (PCA) followed by both statistical and machine learning classification methods, and compared them to a popular machine learning technique, Support Vector Machines (SVM). Both PLS and SVM demonstrate strong utility for proteomic classification problems.     

HitKeeper, a generic software package for hit list management

Background  

The automated annotation of biological sequences (protein, DNA) relies on the computation of hits (predicted features) on the sequences using various algorithms. Public databases of biological sequences provide a wealth of biological "knowledge", for example manually validated annotations (features) that are located on the sequences, but mining the sequence annotations and especially the predicted and curated features requires dedicated tools. Due to the heterogeneity and diversity of the biological information, it is difficult to handle redundancy, frequent updates, taxonomic information and "private" data together with computational algorithms in a common workflow.     

Using an Adjusted Serfling Regression Model to Improve the Early Warning at the Arrival of Peak Timing of Influenza in Beijing

Serfling-type periodic regression models have been widely used to identify and analyse epidemic of influenza. In these approaches, the baseline is traditionally determined using cleaned historical non-epidemic data. However, we found that the previous exclusion of epidemic seasons was empirical, since year-year variations in the seasonal pattern of activity had been ignored. Therefore, excluding fixed ‘epidemic’ months did not seem reasonable. We made some adjustments in the rule of epidemic-period removal to avoid potentially subjective definition of the start and end of epidemic periods. We fitted the baseline iteratively. Firstly, we established a Serfling regression model based on the actual observations without any removals. After that, instead of manually excluding a predefined ‘epidemic’ period (the traditional method), we excluded observations which exceeded a calculated boundary. We then established Serfling regression once more using the cleaned data and excluded observations which exceeded a calculated boundary. We repeated this process until the R² value stopped to increase. In addition, the definitions of the onset of influenza epidemic were heterogeneous, which might make it impossible to accurately evaluate the performance of alternative approaches. We then used this modified model to detect the peak timing of influenza instead of the onset of epidemic and compared this model with traditional Serfling models using observed weekly case counts of influenza-like illness (ILIs), in terms of sensitivity, specificity and lead time. A better performance was observed. In summary, we provide an adjusted Serfling model which may have improved performance over traditional models in early warning at arrival of peak timing of influenza.     

Extracting diagnoses and investigation results from unstructured text in electronic health records by semi-supervised machine learning

Background

Electronic health records are invaluable for medical research, but much of the information is recorded as unstructured free text which is time-consuming to review manually.

Aim

To develop an algorithm to identify relevant free texts automatically based on labelled examples.

Methods

We developed a novel machine learning algorithm, the ‘Semi-supervised Set Covering Machine’ (S3CM), and tested its ability to detect the presence of coronary angiogram results and ovarian cancer diagnoses in free text in the General Practice Research Database. For training the algorithm, we used texts classified as positive and negative according to their associated Read diagnostic codes, rather than by manual annotation. We evaluated the precision (positive predictive value) and recall (sensitivity) of S3CM in classifying unlabelled texts against the gold standard of manual review. We compared the performance of S3CM with the Transductive Vector Support Machine (TVSM), the original fully-supervised Set Covering Machine (SCM) and our ‘Freetext Matching Algorithm’ natural language processor.

Results

Only 60% of texts with Read codes for angiogram actually contained angiogram results. However, the S3CM algorithm achieved 87% recall with 64% precision on detecting coronary angiogram results, outperforming the fully-supervised SCM (recall 78%, precision 60%) and TSVM (recall 2%, precision 3%). For ovarian cancer diagnoses, S3CM had higher recall than the other algorithms tested (86%). The Freetext Matching Algorithm had better precision than S3CM (85% versus 74%) but lower recall (62%).

Conclusions

Our novel S3CM machine learning algorithm effectively detected free texts in primary care records associated with angiogram results and ovarian cancer diagnoses, after training on pre-classified test sets. It should be easy to adapt to other disease areas as it does not rely on linguistic rules, but needs further testing in other electronic health record datasets.     

Taxonomic Classification of Bacterial 16S rRNA Genes Using Short Sequencing Reads: Evaluation of Effective Study Designs

Massively parallel high throughput sequencing technologies allow us to interrogate the microbial composition of biological samples at unprecedented resolution. The typical approach is to perform high-throughout sequencing of 16S rRNA genes, which are then taxonomically classified based on similarity to known sequences in existing databases. Current technologies cause a predicament though, because although they enable deep coverage of samples, they are limited in the length of sequence they can produce. As a result, high-throughout studies of microbial communities often do not sequence the entire 16S rRNA gene. The challenge is to obtain reliable representation of bacterial communities through taxonomic classification of short 16S rRNA gene sequences. In this study we explored properties of different study designs and developed specific recommendations for effective use of short-read sequencing technologies for the purpose of interrogating bacterial communities, with a focus on classification using naïve Bayesian classifiers. To assess precision and coverage of each design, we used a collection of ∼8,500 manually curated 16S rRNA gene sequences from cultured bacteria and a set of over one million bacterial 16S rRNA gene sequences retrieved from environmental samples, respectively. We also tested different configurations of taxonomic classification approaches using short read sequencing data, and provide recommendations for optimal choice of the relevant parameters. We conclude that with a judicious selection of the sequenced region and the corresponding choice of a suitable training set for taxonomic classification, it is possible to explore bacterial communities at great depth using current technologies, with only a minimal loss of taxonomic resolution.     

A quick mind with letters can be a slow mind with natural scenes: individual differences in attentional selection

Background

Most people show a remarkable deficit in reporting the second of two targets (T2) when presented 200–500 ms after the first (T1), reflecting an ‘attentional blink’ (AB). However, there are large individual differences in the magnitude of the effect, with some people, referred to as ‘non-blinkers’, showing no such attentional restrictions.

Methodology/Principal Findings

Here we replicate these individual differences in a task requiring identification of two letters amongst digits, and show that the observed differences in T2 performance cannot be attributed to individual differences in T1 performance. In a second experiment, the generality of the non-blinkers'' superior performance was tested using a task containing novel pictures rather than alphanumeric stimuli. A substantial AB was obtained in non-blinkers that was equivalent to that of ‘blinkers’.

Conclusion/Significance

The results suggest that non-blinkers employ an efficient target selection strategy that relies on well-learned alphabetic and numeric category sets.     

Spectral dictionaries: Integrating de novo peptide sequencing with database search of tandem mass spectra

Database search tools identify peptides by matching tandem mass spectra against a protein database. We study an alternative approach when all plausible de novo interpretations of a spectrum (spectral dictionary) are generated and then quickly matched against the database. We present a new MS-Dictionary algorithm for efficiently generating spectral dictionaries and demonstrate that MS-Dictionary can identify spectra that are missed in the database search. We argue that MS-Dictionary enables proteogenomics searches in six-frame translation of genomic sequences that may be prohibitively time-consuming for existing database search approaches. We show that such searches allow one to correct sequencing errors and find programmed frameshifts.     

Semi-Supervised Morphosyntactic Classification of Old Icelandic

We present IceMorph, a semi-supervised morphosyntactic analyzer of Old Icelandic. In addition to machine-read corpora and dictionaries, it applies a small set of declension prototypes to map corpus words to dictionary entries. A web-based GUI allows expert users to modify and augment data through an online process. A machine learning module incorporates prototype data, edit-distance metrics, and expert feedback to continuously update part-of-speech and morphosyntactic classification. An advantage of the analyzer is its ability to achieve competitive classification accuracy with minimum training data.     

Pfam: A comprehensive database of protein domain families based on seed alignments

Databases of multiple sequence alignments are a valuable aid to protein sequence classification and analysis. One of the main challenges when constructing such a database is to simultaneously satisfy the conflicting demands of completeness on the one hand and quality of alignment and domain definitions on the other. The latter properties are best dealt with by manual approaches, whereas completeness in practice is only amenable to automatic methods. Herein we present a database based on hidden Markov model profiles (HMMs), which combines high quality and completeness. Our database, Pfam, consists of parts A and B. Pfam-A is curated and contains well-characterized protein domain families with high quality alignments, which are maintained by using manually checked seed alignments and HMMs to find and align all members. Pfam-B contains sequence families that were generated automatically by applying the Domainer algorithm to cluster and align the remaining protein sequences after removal of Pfam-A domains. By using Pfam, a large number of previously unannotated proteins from the Caenorhabditis elegans genome project were classified. We have also identified many novel family memberships in known proteins, including new kazal, Fibronectin type III, and response regulator receiver domains. Pfam-A families have permanent accession numbers and form a library of HMMs available for searching and automatic annotation of new protein sequences. Proteins: 28:405–420, 1997. © 1997 Wiley-Liss, Inc.     

A word-oriented approach to alignment validation

MOTIVATION: Multiple sequence alignment at the level of whole proteomes requires a high degree of automation, precluding the use of traditional validation methods such as manual curation. Since evolutionary models are too general to describe the history of each residue in a protein family, there is no single algorithm/model combination that can yield a biologically or evolutionarily optimal alignment. We propose a 'shotgun' strategy where many different algorithms are used to align the same family, and the best of these alignments is then chosen with a reliable objective function. We present WOOF, a novel 'word-oriented' objective function that relies on the identification and scoring of conserved amino acid patterns (words) between pairs of sequences. RESULTS: Tests on a subset of reference protein alignments from BAliBASE showed that WOOF tended to rank the (manually curated) reference alignment highest among 1060 alternative (automatically generated) alignments for a majority of protein families. Among the automated alignments, there was a strong positive relationship between the WOOF score and similarity to the reference alignment. The speed of WOOF and its independence from explicit considerations of three-dimensional structure make it an excellent tool for analyzing large numbers of protein families. AVAILABILITY: On request from the authors.