Ajit Narayanan

Quantum-inspired genetic algorithms

A novel evolutionary computing method-quantum inspired genetic algorithms-is introduced, where concepts and principles of quantum mechanics are used to inform and inspire more efficient evolutionary computing methods. The basic terminology of quantum mechanics is introduced before a comparison is made between a classical genetic algorithm and a quantum inspired method for the travelling salesperson problem. It is informally shown that the quantum inspired genetic algorithm performs better than the classical counterpart for a small domain. The paper concludes with some speculative comments concerning the relationship between quantum inspired genetic algorithms and various complexity classes.

Quantum artificial neural network architectures and components

It is shown by classical simulation and experimentation that quantum artificial neural networks (QUANNs) are more efficient and in some cases more powerful than classical artificial neural networks (CLANNs) for a variety of experimental tasks. This effect is particularly noticeable with larger and more complex domains. The gain in efficiency is achieved with no generalisation loss in most cases. QUANNs are also more powerful than CLANNs, again for some of the tasks examined, in terms of what the network can learn. What is more, it appears that not all components of a QUANN architecture need to to be quantum for these advantages to surface. It is demonstrated that a fully quantum neural network has no advantage over a partly quantum network and may in fact produce worse results. Overall, this work provides a first insight into the expected behaviour of individual components of QUANNs, if and when quantum hardware is ever built, and raises questions about the interface between quantum and classical components of future QUANNs.

Discovering Gene Networks with a Neural-Genetic Hybrid

Recent advances in biology (namely, DNA arrays) allow an unprecedented view of the biochemical mechanisms contained within a cell. However, this technology raises new challenges for computer scientists and biologists alike, as the data created by these arrays is often highly complex. One of the challenges is the elucidation of the regulatory connections and interactions between genes, proteins and other gene products. In this paper, a novel method is described for determining gene interactions in temporal gene expression data using genetic algorithms combined with a neural network component. Experiments conducted on real-world temporal gene expression data sets confirm that the approach is capable of finding gene networks that fit the data. A further repeated approach shows that those genes significantly involved in interaction with other genes can be highlighted and hypothetical gene networks and circuits proposed for further laboratory testing.

Assessing the professional performance of UK doctors: an evaluation of the utility of the General Medical Council patient and colleague questionnaires

Objective: To investigate the utility of the GMC patient and colleague questionnaires in assessing the professional performance of a large sample of UK doctors. Design: Cross-sectional questionnaire surveys. Setting: Range of UK clinical practice settings. Participants: 541 doctors gave preliminary agreement to take part in the study. Responses were received from 13 754 patients attending one of 380 participant doctors, and from 4269 colleagues of 309 participant doctors. Main outcome measures: Questionnaire performance and standardised scores for each doctor derived from patient and colleague responses. Results: Participant doctors were similar to non-participants in respect of age and gender. The patient and colleague questionnaires were acceptable to participants as evidenced by low levels of missing data. One patient questionnaire item seemed to cause confusion for respondents and requires rewording. Both patient and colleague responses were highly skewed towards favourable impressions of doctor performance, with high internal consistency. To achieve acceptable levels of reliability, a minimum of 8 colleague questionnaires and 22 patient questionnaires are required. G coefficients for both questionnaires were comparable with internationally recognised survey instruments of broadly similar intent. Patient and colleague assessments provided complementary perspectives of doctors’ performance. Older doctors had lower patient-derived and colleague-derived scores than younger doctors. Doctors from a mental health trust and doctors providing care in a variety of non-NHS settings had lower patient scores compared with doctors providing care in acute or primary care trust settings. Conclusions: The GMC patient and colleague questionnaires offer a reliable basis for the assessment of professionalism among UK doctors. If used in the revalidation of doctors’ registration, they would be capable of discriminating a range of professional performance among doctors, and potentially identifying a minority whose practice should to subjected to further scrutiny.

Quantum computing for beginners

The paper introduces the basic concepts and principles behind quantum computing and examines in detail Shors (1994) quantum algorithm for factoring very large numbers. Some basic methodological principles and guidelines for constructing quantum algorithms are stated. The aim is not to provide a formal exposition of quantum computing but to identify its novelty and potential use in tackling NP-hard problems.

Genetic algorithms for gene expression analysis

The major problem for current gene expression analysis techniques is how to identify the handful of genes which contribute to a disease from the thousands of genes measured on gene chips (microarrays). The use of a novel neural-genetic hybrid algorithm for gene expression analysis is described here. The genetic algorithm identifies possible gene combinations for classification and then uses the output from a neural network to determine the fitness of these combinations. Normal mutation and crossover operations are used to find increasingly fit combinations. Experiments on artificial and real-world gene expression databases are reported. The results from the algorithm are also explored for biological plausibility and confirm that the algorithm is a powerful alternative to standard data mining techniques in this domain.

Single-layer artificial neural networks for gene expression analysis

Gene expression datasets are being produced in increasing quantities and made available on the web. Several thousands of genes are usually measured for their mRNA expression levels per sample using Affymetrix gene chips and Stanford microarrays, for instance. Such datasets are normally separated into distinct, objectively measured classes, typically disease states or other objectively measured phenotypes. A major problem for current gene expression analysis is, given the disparity between the number of genes measured (typically, thousands) and number of individuals sampled (typically, dozens), how to identify the handful of genes which, individually or in combination, help classify individuals. Previous approaches when faced with the dimensionality of the problem have tended to use unsupervised or supervised techniques that result in smaller clusters of genes, but clusters by themselves do not yield classification rules. This is especially the case with temporal microarray data, which represents the activity of genes within a cell, tissue or organism over time. The expression levels of some genes at a particular time-point can be controlled by the expression levels of other genes at a previous time-point. It is the extraction of these temporal connections within the data that is of great interest to biomolecular scientists and researchers within the pharmaceutical industry. If these so-called gene networks can be found that explain disease inception and progression, drugs can be designed to target specific genes so that the disease either does not progress or is even eradicated from an individual. In this paper we describe novel experiments using single-layer artificial neural networks for modelling both non-temporal (classificatory) and temporal microarray data.

The molecular basis of the chemosensitivity of metastatic cutaneous melanoma to chemotherapy

Background Chemotherapy benefits relatively few patients with cutaneous melanoma. The assessment of tumour chemosensitivity by the ATP-based tumour chemosensitivity assay (ATP-TCA) has shown strong correlation with outcome in cutaneous melanoma, but requires fresh tissue and dedicated laboratory facilities. Aim To examine whether the results of the ATP-TCA correlate with the expression of genes known to be involved in resistance to chemotherapy, based on the hypothesis that the molecular basis of chemosensitivity lies within known drug resistance mechanisms. Method The chemosensitivity of 47 cutaneous melanomas was assessed using the ATP-TCA and correlated with quantitative expression of 93 resistance genes measured by quantitative reverse transcriptase PCR (qRT-PCR) in a Taqman Array after extraction of total RNA from formalin-fixed paraffin-embedded tissue. Results Drugs susceptible to particular resistance mechanisms showed good correlation with genes linked to these mechanisms using signatures of up to 17 genes. Comparison of these signatures for DTIC, treosulfan and cisplatin showed several genes in common. HSP70, at least one human epidermal growth factor receptor, genes involved in apoptosis (IAP2, PTEN) and DNA repair (ERCC1, XPA, XRCC1, XRCC6) were present for these agents, as well as genes involved in the regulation of proliferation (Ki67, p21, p27). The combinations tested included genes represented in the single agent signatures. Conclusions These data suggest that melanoma chemosensitivity is influenced by known resistance mechanisms, including susceptibility to apoptosis. Use of a candidate gene approach may increase understanding of the mechanisms underlying chemosensitivity to drugs active against melanoma and provide signatures with predictive value.

Validation of a multi-source feedback tool for use in general practice.

Feedback from colleagues and patients is a core element of the revalidation process being developed by the General Medical Council. However, there are few feedback tools which have been specifically developed and validated for doctors in primary care. This paper presents data demonstrating the reliability and validity of one such tool. The CFEP360 tool combines feedback from the Colleague Feedback Evaluation Tool (CFET) and the Doctors Interpersonal Skills Questionnaire (DISQ). The analysis of over 10 000 completed questionnaires presented here identifies that colleague feedback is essentially two-dimensional (i.e. clinical and non-clinical skills) and that patient feedback is one-dimensional. However, items from both scales also effectively predict combined global ratings, indicating that colleagues and patients are identifying similar levels of performance as accessed by the feedback. Doctors who receive low feedback scores may require further attention, meaning the feedback potentially has diagnostic value. Reliable feedback on this tool, as indicated by this analysis, requires 14 colleague responses and 25 patient responses, figures comparable to other MSF tools if CFEP360 is to be used for a high stakes performance evaluation and possible revalidation (generalisability statistic G> or =0.80). For lower stakes performance evaluations, such as personal development, responses from 11 colleagues and 16 patients will still return reliable results (G> or =0.70).

Two-Phase EA/k-NN for Feature Selection and Classification in Cancer Microarray Datasets

Efficient and reliable methods that can find a small sample of informative genes amongst thousands are of great importance. In this area, much research is investigating the combination of advanced search strategies (to find subsets of features), and classification methods. We investigate a simple evolutionary algorithm/classifier combination on two microarray cancer datasets, where this combination is applied twice – once for feature selection, and once for further selection and classification. Our contribution are: (further) demonstration that a simple EA/classifier combination is capable of good feature discovery and classification performance with no initial dimensionality reduction; demonstration that a simple repeated EA/k-NN approach is capable of competitive or better performance than methods using more sophisticated preprocessing and classifer methods; new and challenging results on two public datasets with clear explanation of experimental setup; review material on the EA/kNN area; and specific identification of genes that our work suggests are significant regarding colon cancer and prostate cancer.