Learning with symmetric label noise: The importance of being unhinged

Convex potential minimisation is the de facto approach to binary classification. However, Long and Servedio [2008] proved that under symmetric label noise (SLN), minimisation of any convex potential over a linear function class can result in classification performance equivalent to random guessing. This ostensibly shows that convex losses are not SLN-robust. In this paper, we propose a convex, classification-calibrated loss and prove that it is SLN-robust. The loss avoids the Long and Servedio [2008] result by virtue of being negatively unbounded. The loss is a modification of the hinge loss, where one does not clamp at zero; hence, we call it the unhinged loss. We show that the optimal unhinged solution is equivalent to that of a strongly regularised SVM, and is the limiting solution for any convex potential; this implies that strong l2 regularisation makes most standard learners SLN-robust. Experiments confirm the unhinged loss’ SLN-robustness.

Maximising Revenue within the NT Mud Crab Fishery by Enhancing Post-Harvest Survival of Mud Crabs

Identifying the stress factors imposed on mud crab to develop stress minimisation practices for improving survival, hence increasing revenue for the industry.

Gene expression evidence for off-target effects caused by RNA interference-mediated gene silencing of Ubiquitin-63E in the cattle tick Rhipicephalus microplus.

Knowledge of cattle tick (Rhipicephalus (Boophilus) microplus; Acari: Ixodidae) molecular and cellular pathways has been hampered by the lack of an annotated genome. In addition, most of the tick expressed sequence tags (ESTs) available to date consist of similar to 50% unassigned sequences without predicted functions. The most common approach to address this has been the application of RNA interference (RNAi) methods to investigate genes and their pathways. This approach has been widely adopted in tick research despite minimal knowledge of the tick RNAi pathway and double-stranded RNA (dsRNA) uptake mechanisms. A strong knockdown phenotype of adult female ticks had previously been observed using a 594 bp dsRNA targeting the cattle tick homologue for the Drosophila Ubiquitin-63E gene leading to nil or deformed eggs. A NimbleGen cattle tick custom microarray based on the BmiGI.V2 database of R. microplus ESTs was used to evaluate the expression of mRNAs harvested from ticks treated with the tick Ubiquitin-63E 594 bp dsRNA compared with controls. A total of 144 ESTs including TC6372 (Ubiquitin-63E) were down-regulated with 136 ESTs up-regulated following treatment. The results obtained substantiated the knockdown phenotype with ESTs identified as being associated with ubiquitin proteolysis as well as oogenesis, embryogenesis, fatty acid synthesis and stress responses. A bioinformatics analysis was undertaken to predict off-target effects (OTE) resulting from the in silico dicing of the 594 bp Ubiquitin-63E dsRNA which identified 10 down-regulated ESTs (including TC6372) within the list of differentially expressed probes on the microarrays. Subsequent knockdown experiments utilising 196 and 109 bp dsRNAs, and a cocktail of short hairpin RNAs (shRNA) targeting Ubiquitin-63E, demonstrated similar phenotypes for the dsRNAs but nil effect following shRNA treatment. Quantitative reverse transcriptase PCR analysis confirmed differential expression of TC6372 and selected ESTs. Our study demonstrated the minimisation of predicted OTEs in the shorter dsRNA treatments (similar to 100-200 bp) and the usefulness of microarrays to study knockdown phenotypes.

Hendra Virus and Horse Owners – Risk Perception and Management

Hendra virus is a highly pathogenic novel paramyxovirus causing sporadic fatal infection in horses and humans in Australia. Species of fruit-bats (genus Pteropus), commonly known as flying-foxes, are the natural host of the virus. We undertook a survey of horse owners in the states of Queensland and New South Wales, Australia to assess the level of adoption of recommended risk management strategies and to identify impediments to adoption. Survey questionnaires were completed by 1431 respondents from the target states, and from a spectrum of industry sectors. Hendra virus knowledge varied with sector, but was generally limited, with only 13% of respondents rating their level of knowledge as high or very high. The majority of respondents (63%) had seen their state’s Hendra virus information for horse owners, and a similar proportion found the information useful. Fifty-six percent of respondents thought it moderately, very or extremely likely that a Hendra virus case could occur in their area, yet only 37% said they would consider Hendra virus if their horse was sick. Only 13% of respondents stabled their horses overnight, although another 24% said it would be easy or very easy to do so, but hadn’t done so. Only 13% and 15% of respondents respectively had horse feed bins and water points under solid cover. Responses varied significantly with state, likely reflecting different Hendra virus history. The survey identified inconsistent awareness and/or adoption of available knowledge, confusion in relation to Hendra virus risk perception, with both over-and under-estimation of true risk, and lag in the uptake of recommended risk minimisation strategies, even when these were readily implementable. However, we also identified frustration and potential alienation by horse owners who found the recommended strategies impractical, onerous and prohibitively expensive. The insights gained from this survey have broader application to other complex risk-management scenarios.

A comparison of welfare outcomes for weaner and mature Bos indicus bulls surgically or tension band castrated with or without analgesia: 1. Behavioural responses

Tension-band castration of cattle is gaining favour because it is relatively simple to perform and is promoted by retailers of the devices as a humane castration method. Furthermore, retailers encourage delaying castration to exploit the superior growth rates of bulls compared with steers. Two experiments were conducted, under tropical conditions, comparing tension banding and surgical castration of weaner (7–10 months old) and mature (22–25 months old) Bos indicus bulls with and without pain management (ketoprofen or saline injected intramuscularly immediately prior to castration). Welfare outcomes were assessed using a wide range of measures; this paper reports on the behavioural responses of the bulls and an accompanying paper reports on other measures. Behavioural data were collected at intervals by direct observation and continuously via data loggers on the hind leg of the bulls to 4 weeks post-castration. Tension-banded bulls performed less movement in the crush/chute than the surgically castrated bulls during the procedures (weaner: 2.63 vs. 5.69, P < 0.001; mature: 1.00 vs. 5.94; P < 0.001 for tension-band and surgical castration, respectively), indicating that tension banding was less painful then surgical castration during conduct. To 1.5 h post-castration, tension-banded bulls performed significantly (all P < 0.05) more active behavioural responses indicative of pain compared with surgical castrates, e.g., percentage time walking forwards (weaner: 15.0% vs. 8.1%; mature: 22.3% vs. 15.1%), walking backwards (weaner: 4.3% vs. 1.4%; mature: 2.4% vs. 0.5%), numbers of tail movements (weaner: 21.9 vs. 1.4; mature: 51.5 vs. 39.4) and leg movements (weaner: 12.9 vs. 0.9; mature: 8.5 vs. 1.5), respectively. In contrast, surgically castrated bulls performed more immobile behaviours compared with tension-banded bulls (e.g., standing in mature bulls was 56.6% vs. 34.4%, respectively, P = 0.002). Ketoprofen administration appeared effective in moderating pain-related behaviours in the mature bulls from 1.5 to 3 h, e.g., reducing abnormal standing (0.0% vs. 7.7%, P = 0.009) and increasing feeding (12.7% vs. 0.0%, P = 0.048) in NSAID- and saline-treated bulls, respectively. There were few behavioural differences subsequent to 24 h post-castration, but some limited evidence of chronic pain (3–4 weeks post-castration) with both methods. Interpretation, however, was difficult from behaviours alone. Thus, tension banding is less painful than surgical castration during conduct of the procedures and pain-related behavioural responses differ with castration method (active restlessness in response to tension banding and minimisation of movement in response to surgical castration). Ketoprofen administered immediately prior to castration was somewhat effective in reducing pain, particularly in the mature bulls.

Evaluation of packaging films to extend storage life of indigenous Australian vegetables and herbs

A study undertaken in Hervey Bay, Queensland, investigated the potential of creating an indigenous agribusiness opportunity based on the cultivation of indigenous Australian vegetables and herbs. Included were warrigal greens (WG) (Tetragonia tetragonioides), a green leafy vegetable and the herb sea celery (SC) (Apium prostratum); both traditional foods of the indigenous population and highly desirable to chefs wishing to add a unique, indigenous flavour to modern dishes. Packaging is important for shelf life extension and minimisation of postharvest losses in horticultural products. The ability of two packaging films to extend WG and SC shelf life was investigated. These were Antimisted Biaxial Oriented Polypropylene packaging film (BOPP) without perforations and Antifog BOPP Film with microperforations. Weight loss, packaging headspace composition, colour changes, sensory differences and microbial loads of packed WG and SC leaves were monitored to determine the impact of film oxygen transmission rate (OTR) and film water vapour transmission (WVT) on stored product quality. WG and SC were harvested, sanitised, packed and stored at 4°C for 16 days. Results indicated that the OTR and WVT rates of the package film significantly (PKLEINERDAN0.05) influenced the package headspace and weight loss, but did not affect product colour, total bacteria, yeast and mould populations during storage. There was no significant difference (PGROTERDAN0.05) in aroma, appearance, texture and flavour for WG and SC during storage. It was therefore concluded that a shelf life of 16 days at 4°C, where acceptable sensory properties were retained, was achievable for WG and SC in both packaging films.

Construction of a global map of human gene expression : the process, tools and analysis

This thesis studies human gene expression space using high throughput gene expression data from DNA microarrays. In molecular biology, high throughput techniques allow numerical measurements of expression of tens of thousands of genes simultaneously. In a single study, this data is traditionally obtained from a limited number of sample types with a small number of replicates. For organism-wide analysis, this data has been largely unavailable and the global structure of human transcriptome has remained unknown. This thesis introduces a human transcriptome map of different biological entities and analysis of its general structure. The map is constructed from gene expression data from the two largest public microarray data repositories, GEO and ArrayExpress. The creation of this map contributed to the development of ArrayExpress by identifying and retrofitting the previously unusable and missing data and by improving the access to its data. It also contributed to creation of several new tools for microarray data manipulation and establishment of data exchange between GEO and ArrayExpress. The data integration for the global map required creation of a new large ontology of human cell types, disease states, organism parts and cell lines. The ontology was used in a new text mining and decision tree based method for automatic conversion of human readable free text microarray data annotations into categorised format. The data comparability and minimisation of the systematic measurement errors that are characteristic to each lab- oratory in this large cross-laboratories integrated dataset, was ensured by computation of a range of microarray data quality metrics and exclusion of incomparable data. The structure of a global map of human gene expression was then explored by principal component analysis and hierarchical clustering using heuristics and help from another purpose built sample ontology. A preface and motivation to the construction and analysis of a global map of human gene expression is given by analysis of two microarray datasets of human malignant melanoma. The analysis of these sets incorporate indirect comparison of statistical methods for finding differentially expressed genes and point to the need to study gene expression on a global level.

Thin capitalisation – the multinational tax avoidance strategy the budget forgot

We have already seen major amendments to Australia’s tax regime to tackle base erosion and profit shifting (BEPS). Several more significant measures were announced in the federal budget, most notably the diverted profits tax, aimed at multinationals which shift tax to a lower taxing jurisdiction. Yet to date, a very simple tax minimisation strategy has been largely ignored in the ongoing reforms and was ignored in the federal budget. Excessive debt loading is a problem that not been afforded the same attention as other aggressive tax planning strategies adopted by multinationals.

Computer simulation of protein—carbohydrate complexes: application to arabinose-binding protein and pea lectin

The CCEM method (Contact Criteria and Energy Minimisation) has been developed and applied to study protein-carbohydrate interactions. The method uses available X-ray data even on the native protein at low resolution (above 2.4 Å) to generate realistic models of a variety of proteins with various ligands.The two examples discussed in this paper are arabinose-binding protein (ABP) and pea lectin. The X-ray crystal structure data reported on ABP-β-l-arabinose complex at 2.8, 2.4 and 1.7 Å resolution differ drastically in predicting the nature of the interactions between the protein and ligand. It is shown that, using the data at 2.4 Å resolution, the CCEM method generates complexes which are as good as the higher (1.7 Å) resolution data. The CCEM method predicts some of the important hydrogen bonds between the ligand and the protein which are missing in the interpretation of the X-ray data at 2.4 Å resolution. The theoretically predicted hydrogen bonds are in good agreement with those reported at 1.7 Å resolution. Pea lectin has been solved only in the native form at 3 Å resolution. Application of the CCEM method also enables us to generate complexes of pea lectin with methyl-α-d-glucopyranoside and methyl-2,3-dimethyl-α-d-glucopyranoside which explain well the available experimental data in solution.

Computer modelling studies on the mechanism of action of ribonuclease T1

The mechanism of action of ribonuclease (RNase) T1 is still a matter of considerable debate as the results of x-ray, 2-D nmr and site-directed mutagenesis studies disagree regarding the role of the catalytically important residues. Hence computer modelling studies were carried out by energy minimisation of the complexes of RNase T1 and some of its mutants (His40Ala, His40Lys, and Glu58Ala) with the substrate guanyl cytosine (GpC), and of native RNase T1 with the reaction intermediate guanosine 2',3'-cyclic phosphate (G greater than p). The puckering of the guanosine ribose moiety in the minimum energy conformer of the RNase T1-GpC (substrate) complex was found to be O4'-endo and not C3'-endo as in the RNase T1-3'-guanylic acid (inhibitor/product) complex. A possible scheme for the mechanism of action of RNase T1 has been proposed on the basis of the arrangement of the catalytically important amino acid residues His40, Glu58, Arg77, and His92 around the guanosine ribose and the phosphate moiety in the RNase T1-GpC and RNase T1-G greater than p complexes. In this scheme, Glu58 serves as the general base group and His92 as the general acid group in the transphosphorylation step. His40 may be essential for stabilising the negatively charged phosphate moiety in the enzyme-transition state complex.

Studies in biogas technology. Part II. Optimisation of plant dimensions

In this paper, the design basis of the conventional Khadi and Village Industries Commission biogas plants has been elucidated. It has been shown that minimisation of the cost of the gas holder alone leads to the narrow and deep digesters of conventional plants. If instead, the total capital cost of the gas holder plus digester is minimised, the optimisation leads to wide and shallow digesters, which are less expensive. To test this alternative, two prototype plants have been designed, constructed and operated. These plants are not only 25–40% cheaper, but their performance is actually slightly better than the conventional plants.

Molecular Mechanics Studies of Homopolymeric and Mixed Sequence Py.Pu*Pu Triple Helices,

DNA triple helices containing two purine strands and one pyrimidine strand (C.G*G and T.A*A) have been studied, using model building followed by energy minimisation, for different orientations of the third strand resulting from variation in the hydrogen bonding between the Watson-Crick duplex and the third strand and the glycosidic torsion angle in the third strand. Our results show that in the C.G*G case the structure with a parallel orientation of the third strand, resulting from Hoogsteen hydrogen bonds between the third strand and the Watson-Crick duplex, is energetically the most favourable while in the T.A*A case the antiparallel orientation of the third strand, resulting from reverse Hoogsteen hydrogen bonds, is energetically the most favourable. These studies when extended to the mixed sequence triplexes, in which the second strand is a mixture of G and A, correspondingly the third strand is a mixture of G and APT, show that though the parallel orientation is still energetically more favourable, the antiparallel orientation becomes energetically comparable with an increasing number of thymines in the third strand. Structurally, for the mixed triplexes containing G and T in the third strand, it is seen that the basepair non-isomorphism between the C.G*G and the T.A*T triplets can be overcome with some changes in the base pair parameters without much distortion of either the backbone or the hydrogen bonds.

Structural Polymorphism in d(T).d(A)*d(T) Triple Helices

DNA triple helices containing two thymine strands and one adenine strand have been studied, using model building followed by energy minimisation, for different orientations of the third strand resulting from variation in the hydrogen bonding between the Watson-Crick duplex and the third strand and the glycosidic torsion angle in the third strand. Our results show that the structure with a parallel orientation of the third strand, in which the third strand base forms Hoogsteen hydrogen bonds with the adenine base in the Watson-Crick duplex, is energetically the most favourable. An antiparallel orientation of the third strand is also possible, in which the third strand base hydrogen bonds to both the bases in the Watson-Crick duplex. This structure is energetically comparable to the parallel structure. For the parallel triplex a 200ps molecular dynamics simulation starting from two different starting structures indicates that at 300K significant structural heterogeneity exists in this tripler structure. The results are compared with existing structural data on this class of triplexes derived from theoretical and NMR techniques.

Structural characterisation of a uracil containing hairpin DNA by NMR and molecular dynamics

Three-dimensional (3D) structure of a hairpin DNA d-CTAGAGGATCCTTTUGGATCCT (22mer; abbreviated as U4-hairpin), which has a uracil nucleotide unit at the fourth position from the 5' end of the tetra-loop has been solved by NMR spectroscopy. The H-1 resonances of this hairpin have been assigned almost completely. NMR restrained molecular dynamics and energy minimisation procedures have been used to describe the 3D structure of the U4 hairpin. This study establishes that the stem of the hairpin adopts a right handed B-DNA conformation while the T-12 and U-15 nucleotide stack upon 3' and 5' ends of the stem, respectively. Further, T-14 stacks upon both T-12 and U-15 while T-13 partially stacks upon T-14. Very weak stacking interaction is observed between T-13 and T-12. All the individual nucleotide bases adopt 'anti' conformation with respect to their sugar moiety. The turning phosphate in the loop is located between T-13 and T-14. The stereochemistry of U-15 mimics the situation where uracil would stack in a B-DNA conformation. This could be the reason as to why the U4-hairpin is found to be the best substrate for its interaction with uracil DNA glycosylase (UDG) compared to the other substrates in which the uracil is at the first, second and third positions of the tetra-loop from its 5' end, as reported previously.

An analog MVUE for a wireless sensor network

An analog minimum-variance unbiased estimator(MVUE) over an asymmetric wireless sensor network is studied.Minimisation of variance is cast into a constrained non-convex optimisation problem. An explicit algorithm that solves the problem is provided. The solution is obtained by decomposing the original problem into a finite number of convex optimisation problems with explicit solutions. These solutions are then juxtaposed together by exploiting further structure in the objective function.