Integrating Background Knowledge Research Articles

Integration of background knowledge for automatic detection of inconsistencies in gene ontology annotation.

Biological background knowledge plays an important role in the manual quality assurance (QA) of biological database records. One such QA task is the detection of inconsistencies in literature-based Gene Ontology Annotation (GOA). This manual verification ensures the accuracy of the GO annotations based on a comprehensive review of the literature used as evidence, Gene Ontology (GO) terms, and annotated genes in GOA records. While automatic approaches for the detection of semantic inconsistencies in GOA have been developed, they operate within predetermined contexts, lacking the ability to leverage broader evidence, especially relevant domain-specific background knowledge. This paper investigates various types of background knowledge that could improve the detection of prevalent inconsistencies in GOA. In addition, the paper proposes several approaches to integrate background knowledge into the automatic GOA inconsistency detection process. We have extended a previously developed GOA inconsistency dataset with several kinds of GOA-related background knowledge, including GeneRIF statements, biological concepts mentioned within evidence texts, GO hierarchy and existing GO annotations of the specific gene. We have proposed several effective approaches to integrate background knowledge as part of the automatic GOA inconsistency detection process. The proposed approaches can improve automatic detection of self-consistency and several of the most prevalent types of inconsistencies. https://github.com/jiyuc/de-inconsistency.

Less (Learning How to Learn) &[prop] More (Learning): Teaching structure‐function with 3D‐printed models in large, undergraduate biochemistry classes

For students to master core biochemical concepts, they must integrate background knowledge from previous courses in biology, chemistry, physics, and math. Students are often inadequately prepared to make the logical jump from chemical structure representations to biochemical macromolecules, leading to multiple misconceptions about structure-function relationships (Villafañe et al., 2011). These misconceptions often lead students to inaccurately describe basic structural details of macromolecules even after their biochemistry education (Harle and Towns, 2013). Biochemistry educators can bridge this gap, and multiple routes have been demonstrated. However, most of these take significant course development. Thus, there is a need for simple, developed structure-function instructional modules scaffolded for students and instructors. Our team of biochemistry educators, 3D-printing experts, and professional assessment designers collectively developed six interactive learning modules teaching structure-function relationships though cheap, 3D-printed macromolecular models. There first three modules cover protein structure, and begin with students connecting amino acids to physically demonstrate phi and psi angles, through secondary, tertiary and quaternary protein structures, and end with the dynamics of product inhibition. The second three modules cover nucleic acids, and ask students to investigate chemical and structural differences between major and minor grooves, transcription-factor/DNA interactions and supercoiling dynamics. Pre- post-module assessments of intervention or control classes show students working with the models made dramatic and significant learning gains over control students. Pre- post-course assessments show the knowledge is retained. We provide all the materials necessary for interested instructors to modify the modules to work in their courses. We also provide an all-virtual module that can be deployed during the pandemic. Harle M, Towns MH (2013) Students' understanding of primary and secondary protein structure: Drawing secondary protein structure reveals student understanding better than simple recognition of structures. Biochemistry and Molecular Biology Education 41: 369-376 Villafañe SM, Bailey CP, Loertscher J, Minderhout V, Lewis JE (2011) Development and analysis of an instrument to assess student understanding of foundational concepts before biochemistry coursework*. Biochemistry and Molecular Biology Education 39: 102-109

Quantitative Variants of Language Equations and their Applications to Description Logics

Unification in description logics (DLs) has been introduced as a novel inference service that can be used to detect redundancies in ontologies, by finding different concepts that may potentially stand for the same intuitive notion. Together with the special case of matching, they were first investigated in detail for the DL $${\mathcal{FL}}_0$$ , where these problems can be reduced to solving certain language equations. In this thesis, we extend this service in two directions. In order to increase the recall of this method for finding redundancies, we introduce and investigate the notion of approximate unification, which basically finds pairs of concepts that “almost” unify, in order to account for potential small modelling errors. The meaning of “almost” is formalized using distance measures between concepts. We show that approximate unification in $${\mathcal{FL}}_0$$ can be reduced to approximately solving language equations, and devise algorithms for solving the latter problem for particular distance measures. Furthermore, we make a first step towards integrating background knowledge, formulated in so-called TBoxes, by investigating the special case of matching in the presence of TBoxes of different forms. We acquire a tight complexity bound for the general case, while we prove that the problem becomes easier in a restricted setting. To achieve these bounds, we take advantage of an equivalence characterization of $${\mathcal{FL}}_0$$ concepts that is based on formal languages. Even though our results on the approximate setting cannot deal with TBoxes yet, we prepare the framework that future research can build on.

Structured Bayesian Networks: From Inference to Learning with Routes

Structured Bayesian networks (SBNs) are a recently proposed class of probabilistic graphical models which integrate background knowledge in two forms: conditional independence constraints and Boolean domain constraints. In this paper, we propose the first exact inference algorithm for SBNs, based on compiling a given SBN to a Probabilistic Sentential Decision Diagram (PSDD). We further identify a tractable subclass of SBNs, which have PSDDs of polynomial size. These SBNs yield a tractable model of route distributions, whose structure can be learned from GPS data, using a simple algorithm that we propose. Empirically, we demonstrate the utility of our inference algorithm, showing that it can be an order-ofmagnitude more efficient than more traditional approaches to exact inference. We demonstrate the utility of our learning algorithm, showing that it can learn more accurate models and classifiers from GPS data.

The Right Kind of Guesswork

The Right Kind of Guesswork

The Interplay between Text-based Vocabulary Size and Reading Comprehension of Turkish EFL Learners

Recently, there has been a burgeoning interest in relationship between reading comprehension and vocabulary knowledge, and this emanates from fact that academic achievement is closely related to reading performance (Adamson, 1993; Collier, 1989). As might be expected, best way to learn new vocabulary items is considered to read, and knowing extensive vocabulary is a prerequisite to understand a text (Eskey, 2005). Therefore, it is necessary to scrutinize breadth of a person's vocabulary as a predictor of reading comprehension on a regular basis to reach a consensus about amount of vocabulary needed by an L2 learner for a reasonable comprehension of expository texts. However, although there have been several studies evaluating relationship between reading comprehension and vocabulary knowledge in general, these studies do not include Turkish learners of English as participants whose native language is not etymologically related to English. Scholars, on other hand, acknowledge that English as a foreign language (EFL) learners feel burden of reading in an L2 twice as much as their counterparts, L1 readers, do, and success is difficult to come by without being a skilled reader (Carrell & Grabe, 2002). For that reason, present article aims to investigate a perennial concern, that is, relationship between vocabulary coverage and reading comprehension of a group of Turkish EFL learners because first reading is an indispensable skill for academic achievement and second reading comprehension is directly linked to learners' vocabulary knowledge. After a brief summary about reading comprehension and vocabulary knowledge as a reciprocal process, article aims to discuss what percent of vocabulary items should be known in an expository text to comprehend it, and corresponding vocabulary size will be discussed by analyzing sizes suggested in studies of Nation (2006), and Laufer and Ravenhorst-Kalovski (2010).A Reciprocal Process: Reading Comprehension and Vocabulary SizeReading is delineated as the process of receiving and interpreting information encoded in language form via medium of print (Urquhart & Weir, 1998, p. 22). As stated by Linan-Thompson and Vaughn (2007), and Grabe and Stoller (2002), reading comprehension is main purpose for reading, and this purpose underlies and supports most of other purposes for reading. However, since reading involves cognitive processes, reading comprehension is an invisible concept that can only be inferred (Bernhardt, 2011). Moreover, reading comprehension presents some challenges for learners as many students consider reading a boring and difficult task. When complexity of reading is considered together with its purposes and properties, it becomes clear that reading is complex for both teaching and learning.In addition to complexity of reading, it is unarguably clear that relationship between vocabulary size and reading comprehension is reciprocal. While some scholars focus on effects of vocabulary size on reading comprehension in their studies, some others study effects of reading comprehension on vocabulary size (Eskey, 2005; Hu & Nation, 2000; Nation, 2001; Nation & Angell, 2006). According to Rumelhart (1977) and Stanovich (1980), in evidence-based reading models, bottom-up processes such as word recognition and lexical access go hand in hand with top-down processes such as integrating background knowledge and processing strategies. Readers need automaticity in both word recognition and lexical access (Walter, 2003). From a lexical perspective, Anderson (2009) and DeKeyser (2007) summarize this long learning process as a path from understanding a word's meaning to learning a major meaning of a word, and then learning many aspects of a word's meaning and use. Therefore, faster a reader recognizes a word, which is linked to learners' vocabulary knowledge and automaticity, better reading comprehension will take place. …

Integrative relational machine-learning for understanding drug side-effect profiles

BackgroundDrug side effects represent a common reason for stopping drug development during clinical trials. Improving our ability to understand drug side effects is necessary to reduce attrition rates during drug development as well as the risk of discovering novel side effects in available drugs. Today, most investigations deal with isolated side effects and overlook possible redundancy and their frequent co-occurrence.ResultsIn this work, drug annotations are collected from SIDER and DrugBank databases. Terms describing individual side effects reported in SIDER are clustered with a semantic similarity measure into term clusters (TCs). Maximal frequent itemsets are extracted from the resulting drug x TC binary table, leading to the identification of what we call side-effect profiles (SEPs). A SEP is defined as the longest combination of TCs which are shared by a significant number of drugs. Frequent SEPs are explored on the basis of integrated drug and target descriptors using two machine learning methods: decision-trees and inductive-logic programming. Although both methods yield explicit models, inductive-logic programming method performs relational learning and is able to exploit not only drug properties but also background knowledge. Learning efficiency is evaluated by cross-validation and direct testing with new molecules. Comparison of the two machine-learning methods shows that the inductive-logic-programming method displays a greater sensitivity than decision trees and successfully exploit background knowledge such as functional annotations and pathways of drug targets, thereby producing rich and expressive rules. All models and theories are available on a dedicated web site.ConclusionsSide effect profiles covering significant number of drugs have been extracted from a drug ×side-effect association table. Integration of background knowledge concerning both chemical and biological spaces has been combined with a relational learning method for discovering rules which explicitly characterize drug-SEP associations. These rules are successfully used for predicting SEPs associated with new drugs.

Patterns of Literacy among U.S. Students

How well do U.S. students read? In this article, Sean Reardon, Rachel Valentino, and Kenneth Shores rely on studies using data from national and international literacy assessments to answer this question. In part, the answer depends on the specific literacy skills assessed. The authors show that almost all U.S. students can "read" by third grade, if reading is defined as proficiency in basic procedural word-reading skills. But reading for comprehension--integrating background knowledge and contextual information to make sense of a text--requires a set of knowledge-based competencies in addition to word-reading skills. By the standards used in various large-scale literacy assessments, only about a third of U.S. students in middle school possess the knowledge-based competencies to "read" in this more comprehensive sense. This low level of literacy proficiency does not appear to be a result of declining performance over time. Literacy skills of nine-year-olds in the United States have increased modestly over the past forty years, while the skills of thirteen- and seventeen-year-olds have remained relatively flat. Literacy skills vary considerably among students, however. For example, the literacy skills of roughly 10 percent of seventeen-year-olds are at the level of the typical nine-year-old. This variation is patterned in part by race, ethnicity, and socioeconomic background. Black and Hispanic students enter high school with average literacy skills three years behind those of white and Asian students; students from low-income families enter high school with average literacy skills five years behind those of high-income students. These are gaps that no amount of remedial instruction in high school is likely to eliminate. And while the racial and ethnic disparities are smaller than they were forty to fifty years ago, socioeconomic disparities in literacy skills are growing. Nor is the low level of literacy skills particularly a U.S. phenomenon. On international comparisons, American students perform modestly above average compared with those in other developed countries (and well above average among a larger set of countries). Moreover, there is no evidence that U.S. students lose ground relative to those in other countries during the middle school years. Thus, although literacy skills in the United States are lower than needed to meet the demands of modern society, the same is true in most other developed countries.

Integrating background knowledge from internet databases into predictive toxicology models

While data integration for data analysis has been investigated extensively in biological applications, it has not yet been so much the focus in computational chemistry and quantitative structure–activity relationship (QSAR) research. With the availability and growing number of chemical databases on the web, such data integration efforts become an intriguing possibility (and, in fact, a necessity). In this paper, we take a first step towards the following vision and scenario for predictive toxicology applications. Given a new structure to be predicted, the first step would be to gather (integrate) all relevant information from internet databases for the structure itself, and all structures with available information for the endpoint of interest. In a second step, the collected information is combined statistically into a prediction of the new structure. We simulate this scenario with three endpoints (data sets) from the DSSTox database and collect information from three public chemical databases: PubChem, ChemBank and Sigma-Aldrich. In the experiments, we investigate whether the addition of background knowledge from the three databases can improve predictive performance (over using chemical structure alone) in a statistically significant way. For this purpose, we define groups of features (belonging together from an application point of view) from the three databases, and perform a variant of forward selection to include these feature groups in a prediction model. Our experiments show that the integration of background knowledge from internet databases can significantly improve prediction performance, especially for regression tasks.

Collaborative clustering with background knowledge

The aim of collaborative clustering is to make different clustering methods collaborate, in order to reach at an agreement on the partitioning of a common dataset. As different clustering methods can produce different partitioning of the same dataset, finding a consensual clustering from these results is often a hard task. The collaboration aims to make the methods agree on the partitioning through a refinement of their results. This process tends to make the results more similar. In this paper, after the introduction of the collaboration process, we present different ways to integrate background knowledge into it. Indeed, in recent years, the integration of background knowledge in clustering algorithms has been the subject of a lot of interest. This integration often leads to an improvement of the quality of the results. We discuss how such integration in the collaborative process is beneficial and we present experiments in which background knowledge is used to guide collaboration.

Using literature-based discovery to identify disease candidate genes

We present BITOLA, an interactive literature-based biomedical discovery support system. The goal of this system is to discover new, potentially meaningful relations between a given starting concept of interest and other concepts, by mining the bibliographic database MEDLINE. To make the system more suitable for disease candidate gene discovery and to decrease the number of candidate relations, we integrate background knowledge about the chromosomal location of the starting disease as well as the chromosomal location of the candidate genes from resources such as LocusLink and Human Genome Organization (HUGO). BITOLA can also be used as an alternative way of searching the MEDLINE database. The system is available at http://www.mf.uni-lj.si/bitola/.

Interactive machine learning: letting users build classifiers

According to standard procedure, building a classifier using machine learning is a fully automated process that follows the preparation of training data by a domain expert. In contrast, interactive machine learning engages users in actually generating the classifier themselves. This offers a natural way of integrating background knowledge into the modelling stage—as long as interactive tools can be designed that support efficient and effective communication. This paper shows that appropriate techniques can empower users to create models that compete with classifiers built by state-of-the-art learning algorithms. It demonstrates that users—even users who are not domain experts—can often construct good classifiers, without any help from a learning algorithm, using a simple two-dimensional visual interface. Experiments on real data demonstrate that, not surprisingly, success hinges on the domain: if a few attributes can support good predictions, users generate accurate classifiers, whereas domains with many high-order attribute interactions favour standard machine learning techniques. We also present an artificial example where domain knowledge allows an “expert user” to create a much more accurate model than automatic learning algorithms. These results indicate that our system has the potential to produce highly accurate classifiers in the hands of a domain expert who has a strong interest in the domain and therefore some insights into how to partition the data. Moreover, small expert-defined models offer the additional advantage that they will generally be more intelligible than those generated by automatic techniques.

Toward Evidence-Based Medical Statistics. 1: The P Value Fallacy

An important problem exists in the interpretation of modern medical research data: Biological understanding and previous research play little formal role in the interpretation of quantitative results. This phenomenon is manifest in the discussion sections of research articles and ultimately can affect the reliability of conclusions. The standard statistical approach has created this situation by promoting the illusion that conclusions can be produced with certain "error rates," without consideration of information from outside the experiment. This statistical approach, the key components of which are P values and hypothesis tests, is widely perceived as a mathematically coherent approach to inference. There is little appreciation in the medical community that the methodology is an amalgam of incompatible elements, whose utility for scientific inference has been the subject of intense debate among statisticians for almost 70 years. This article introduces some of the key elements of that debate and traces the appeal and adverse impact of this methodology to the P value fallacy, the mistaken idea that a single number can capture both the long-run outcomes of an experiment and the evidential meaning of a single result. This argument is made as a prelude to the suggestion that another measure of evidence should be used--the Bayes factor, which properly separates issues of long-run behavior from evidential strength and allows the integration of background knowledge with statistical findings.