Michael Bain

iHMMune-align

MOTIVATION Immunoglobulin heavy chain (IGH) genes in mature B lymphocytes are the result of recombination of IGHV, IGHD and IGHJ germline genes, followed by somatic mutation. The correct identification of the germline genes that make up a variable VH domain is essential to our understanding of the process of antibody diversity generation as well as to clinical investigations of some leukaemias and lymphomas. RESULTS We have developed iHMMune-align, an alignment program that uses a hidden Markov model (HMM) to model the processes involved in human IGH gene rearrangement and maturation. The performance of iHMMune-align was compared to that of other immunoglobulin gene alignment utilities using both clonally related and randomly selected IGH sequences. This evaluation suggests that iHMMune-align provides a more accurate identification of component germline genes than other currently available IGH gene characterization programs. AVAILABILITY iHMMune-align cross-platform Java executable and web interface are freely available to academic users and can be accessed at http://www.emi.unsw.edu.au/~ihmmune/.

Experimental comparison of human and machine learning formalisms

In this paper we describe the results of a set of experiments in which we compared the learning performance of human and machine learning agents. The problem involved the learning of a concept description for deciding on the legality of positions within the chess endgame King and Rook against King. Various amounts of background knowledge were made available to each learning agent. We concluded that the ability to produce high performance in this domain was almost entirely dependent on the ability to express first-order predicate relationships.

Collaborative filtering for people to people recommendation in social networks

Predicting people other people may like has recently become an important task in many online social networks. Traditional collaborative filtering approaches are popular in recommender systems to effectively predict user preferences for items. However, in online social networks people have a dual role as both “users” and “items”, e.g., both initiating and receiving contacts. Here the assumption of active users and passive items in traditional collaborative filtering is inapplicable. In this paper we propose a model that fully captures the bilateral role of user interactions within a social network and formulate collaborative filtering methods to enable people to people recommendation. In this model users can be similar to other users in two ways – either having similar “taste” for the users they contact, or having similar “attractiveness” for the users who contact them. We develop SocialCollab, a novel neighbour-based collaborative filtering algorithm to predict, for a given user, other users they may like to contact, based on user similarity in terms of both attractiveness and taste. In social networks this goes beyond traditional, merely taste-based, collaborative filtering for item selection. Evaluation of the proposed recommender system on datasets from a commercial online social network show improvements over traditional collaborative filtering.

Inductive Construction of Ontologies from Formal Concept Analysis

We present an approach to representation and structuring of theories and ontologies based on a formalism of propositional logic programs. Formal concept analysis is adopted to identify structure in theories. This structure, in the form of conjunctive concepts and the relations between them, is used for representation change in theories based on feature construction and iterative program transformation. Ontologies are represented as sets of propositional definite clauses containing named concepts having a superclass–subclass relationship derived from a concept lattice built using formal concept analysis. Logic programming methods are used to incrementally construct and revise such ontologies. An information compression measure is used to guide the operation of structuring theories and ontologies. The clauses defining the ontology are proved to preserve the relationships which hold between formal concepts in the concept lattice. This framework enables inheritance inference not possible from the structured theories alone. Experimental results are presented from an application to a sample of descriptions of computer science academics’ research interests and a reconstruction experiment on randomly generated theories with added noise.

Interaction-based collaborative filtering methods for recommendation in online dating

We consider the problem of developing a recommender system for suggesting suitable matches in an online dating web site. The main problem to be solved is that matches must be highly personalized. Moreover, in contrast to typical product recommender systems, it is unhelpful to recommend popular items: matches must be extremely specific to the tastes and interests of the user, but it is difficult to generate such matches because of the two way nature of the interactions (user initiated contacts may be rejected by the recipient). In this paper, we show that collaborative filtering based on interactions between users is a viable approach in this domain. We propose a number of new methods and metrics to measure and predict potential improvement in user interaction success, which may lead to increased user satisfaction with the dating site. We use these metrics to rigorously evaluate the proposed methods on historical data collected from a commercial online dating web site. The evaluation showed that, had users been able to follow the top 20 recommendations of our best method, their success rate would have improved by a factor of around 2.3.

Reconsidering the human immunoglobulin heavy-chain locus: 1. An evaluation of the expressed human IGHD gene repertoire.

We have used a bioinformatics approach to evaluate the completeness and functionality of the reported human immunoglobulin heavy-chain IGHD gene repertoire. Using the hidden Markov-model-based iHMMune-align program, 1,080 relatively unmutated heavy-chain sequences were aligned against the reported repertoire. These alignments were compared with alignments to 1,639 more highly mutated sequences. Comparisons of the frequencies of gene utilization in the two databases, and analysis of features of aligned IGHD gene segments, including their length, the frequency with which they appear to mutate, and the frequency with which specific mutations were seen, were used to determine the reliability of alignments to the less commonly seen IGHD genes. Analysis demonstrates that IGHD4-23 and IGHD5-24, which have been reported to be open reading frames of uncertain functionality, are represented in the expressed gene repertoire; however, the functionality of IGHD6-25 must be questioned. Sequence similarities make the unequivocal identification of members of the IGHD1 gene family problematic, although all genes except IGHD1-14*01 appear to be functional. On the other hand, reported allelic variants of IGHD2-2 and of the IGHD3 gene family appear to be nonfunctional, very rare, or nonexistent. Analysis also suggests that the reported repertoire is relatively complete, although one new putative polymorphism (IGHD3-10*p03) was identified. This study therefore confirms a surprising lack of diversity in the available IGHD gene repertoire, and restriction of the germline sequence databases to the functional set described here will substantially improve the accuracy of IGHD gene alignments and therefore the accuracy of analysis of the V–D–J junction.

Reciprocal and heterogeneous link prediction in social networks

Link prediction is a key technique in many applications in social networks, where potential links between entities need to be predicted. Conventional link prediction techniques deal with either homogeneous entities, e.g., people to people, item to item links, or non-reciprocal relationships, e.g., people to item links. However, a challenging problem in link prediction is that of heterogeneous and reciprocal link prediction, such as accurate prediction of matches on an online dating site, jobs or workers on employment websites, where the links are reciprocally determined by both entities that heterogeneously belong to disjoint groups. The nature and causes of interactions in these domains makes heterogeneous and reciprocal link prediction significantly different from the conventional version of the problem. In this work, we address these issues by proposing a novel learnable framework called ReHeLP , which learns heterogeneous and reciprocal knowledge from collaborative information and demonstrate its impact on link prediction. Evaluation on a large commercial online dating dataset shows the success of the proposed method and its promise for link prediction.

Instructional Support for Teachers and Guided Feedback for Students in an Adaptive eLearning Environment

Adaptive Tutorials are Intelligent Tutoring Systems (ITS) in which students typically interact with a simulation towards a task-goal while being guided and remediated. Adaptive Tutorials can exhibit different kinds of feedback: students are given guidance based on their interaction, and teachers can also receive feedback on their own authoring choices to drive reflection and content adaptation. This paper will discuss different types of feedback which students are given within Adaptive Tutorials and their pedagogical utility. It will then look at providing teachers with timely support and feedback as well as post-hoc solution traces, to improve student learning. We suggest a refined ITS design and development lifecycle to better support teachers in the design of good pedagogy. We provide this in the form of the Adaptive eLearning Platform, which is a tool to support teachers in the adaptation of tutorials and the creation of customized student feedback. This in turn can result in an improved student experience, with more guidance and better learning.

People-to-People recommendation using multiple compatible subgroups

People-to-people recommendation aims at suggesting suitable matches to people in a way that increases the likelihood of a positive interaction. This problem is more difficult than conventional item-to-people recommendation since the preferences of both parties need to be taken into account. Previously we proposed a profile-based recommendation method that first uses compatible subgroup rules to select a single best attribute value for each corresponding value of the user, then combines these attribute value pairs into a rule that determines the recommendations. Though this method produces a significant improvement in the probability of an interaction being successful, it has two significant limitations: (i) by considering only single matching attribute values the method ignores cases where different attribute values are closely related, missing potential candidates, and (ii) when ranking candidates for recommendation the method does not consider individual behaviour. This paper addresses these two issues, showing how multiple attributes can be used with compatible subgroup rules and individual reply rates used for ranking candidates. Our experimental results show that the new approach significantly improves the probability of an interaction being successful compared to our previous approach.

Knowledge-Guided identification of petri net models of large biological systems

To date, the most expressive, and understandable dynamic models of biological systems identified by ILP have employed qualitative differential equations, or QDEs. The QDE representation provides a direct and simple abstraction of quantitative ODEs. However, the representation does have several limitations, including the generation of spurious behaviour in simulation, and a lack of methods for handling concurrency, quantitative information or stochasticity. These issues are largely absent in the long-established qualitative representation of Petri nets. A flourishing area of Petri net models for biological systems now exists, which has almost entirely been concerned with hand-crafted models. In this paper we show that pure and extended Petri nets can be represented as special cases of systems in which transitions are defined using a combination of logical constraints and constraints on linear terms. Results from a well-known combinatorial algorithm for identifying pure Petri nets from data and from the ILP literature on inverting entailment form the basis of constructing a maximal set of such transition constraints given data and background knowledge. An ILP system equipped with a constraint solver is then used to determine the smallest subset of transition constraints that are consistent with the data. This has several advantages over using a specialised Petri net learner for biological system identification, most of which arise from the use of background knowledge. As a result: (a) search-spaces can be constrained substantially using semantic and syntactic constraints; (b) we can perform the hierarchical identification of Petri models of large systems by re-use of well-established network models; and (c) we can use a combination of abduction and data-based justification to hypothesize missing parts of a Petri net. We demonstrate these advantages on well-known metabolic and signalling networks.