A systematic review of the applications of artificial intelligence and machine learning in autoimmune diseases.

Autoimmune diseases are chronic, multifactorial conditions. Through machine learning (ML), a branch of the wider field of artificial intelligence, it is possible to extract patterns within patient data, and exploit these patterns to predict patient outcomes for improved clinical management. Here, we surveyed the use of ML methods to address clinical problems in autoimmune disease. A systematic review was conducted using MEDLINE, embase and computers and applied sciences complete databases. Relevant papers included "machine learning" or "artificial intelligence" and the autoimmune diseases search term(s) in their title, abstract or key words. Exclusion criteria: studies not written in English, no real human patient data included, publication prior to 2001, studies that were not peer reviewed, non-autoimmune disease comorbidity research and review papers. 169 (of 702) studies met the criteria for inclusion. Support vector machines and random forests were the most popular ML methods used. ML models using data on multiple sclerosis, rheumatoid arthritis and inflammatory bowel disease were most common. A small proportion of studies (7.7% or 13/169) combined different data types in the modelling process. Cross-validation, combined with a separate testing set for more robust model evaluation occurred in 8.3% of papers (14/169). The field may benefit from adopting a best practice of validation, cross-validation and independent testing of ML models. Many models achieved good predictive results in simple scenarios (e.g. classification of cases and controls). Progression to more complex predictive models may be achievable in future through integration of multiple data types.

GenePy - a score for estimating gene pathogenicity in individuals using next-generation sequencing data.

BACKGROUND: Next-generation sequencing is revolutionising diagnosis and treatment of rare diseases, however its application to understanding common disease aetiology is limited. Rare disease applications binarily attribute genetic change(s) at a single locus to a specific phenotype. In common diseases, where multiple genetic variants within and across genes contribute to disease, binary modelling cannot capture the burden of pathogenicity harboured by an individual across a given gene/pathway. We present GenePy, a novel gene-level scoring system for integration and analysis of next-generation sequencing data on a per-individual basis that transforms NGS data interpretation from variant-level to gene-level. This simple and flexible scoring system is intuitive and amenable to integration for machine learning, network and topological approaches, facilitating the investigation of complex phenotypes. RESULTS: Whole-exome sequencing data from 508 individuals were used to generate GenePy scores. For each variant a score is calculated incorporating: i) population allele frequency estimates; ii) individual zygosity, determined through standard variant calling pipelines and; iii) any user defined deleteriousness metric to inform on functional impact. GenePy then combines scores generated for all variants observed into a single gene score for each individual. We generated a matrix of ~ 14,000 GenePy scores for all individuals for each of sixteen popular deleteriousness metrics. All per-gene scores are corrected for gene length. The majority of genes generate GenePy scores < 0.01 although individuals harbouring multiple rare highly deleterious mutations can accumulate extremely high GenePy scores. In the absence of a comparator metric, we examine GenePy performance in discriminating genes known to be associated with three common, complex diseases. A Mann-Whitney U test conducted on GenePy scores for this positive control gene in cases versus controls demonstrates markedly more significant results (p = 1.37 × 10- 4) compared to the most commonly applied association tool that combines common and rare variation (p = 0.003). CONCLUSIONS: Per-gene per-individual GenePy scores are intuitive when assessing genetic variation in individual patients or comparing scores between groups. GenePy outperforms the currently accepted best practice tools for combining common and rare variation. GenePy scores are suitable for downstream data integration with transcriptomic and proteomic data that also report at the gene level.

Classification of Paediatric Inflammatory Bowel Disease using Machine Learning.

Paediatric inflammatory bowel disease (PIBD), comprising Crohn's disease (CD), ulcerative colitis (UC) and inflammatory bowel disease unclassified (IBDU) is a complex and multifactorial condition with increasing incidence. An accurate diagnosis of PIBD is necessary for a prompt and effective treatment. This study utilises machine learning (ML) to classify disease using endoscopic and histological data for 287 children diagnosed with PIBD. Data were used to develop, train, test and validate a ML model to classify disease subtype. Unsupervised models revealed overlap of CD/UC with broad clustering but no clear subtype delineation, whereas hierarchical clustering identified four novel subgroups characterised by differing colonic involvement. Three supervised ML models were developed utilising endoscopic data only, histological only and combined endoscopic/histological data yielding classification accuracy of 71.0%, 76.9% and 82.7% respectively. The optimal combined model was tested on a statistically independent cohort of 48 PIBD patients from the same clinic, accurately classifying 83.3% of patients. This study employs mathematical modelling of endoscopic and histological data to aid diagnostic accuracy. While unsupervised modelling categorises patients into four subgroups, supervised approaches confirm the need of both endoscopic and histological evidence for an accurate diagnosis. Overall, this paper provides a blueprint for ML use with clinical data.

Toward stem cell systems biology: from molecules to networks and landscapes.

The last few years have seen significant advances in our understanding of the molecular mechanisms of stem-cell-fate specification. New and emerging high-throughput techniques, as well as increasingly accurate loss-of-function perturbation techniques, are allowing us to dissect the interplay among genetic, epigenetic, proteomic, and signaling mechanisms in stem-cell-fate determination with ever-increasing fidelity (Boyer et al. 2005, 2006; Ivanova et al. 2006; Loh et al. 2006; Cole et al. 2008; Jiang et al. 2008; Johnson et al. 2008; Kim et al. 2008; Liu et al. 2008; Marson et al. 2008; Mathur et al. 2008). Taken together, recent reports using these new techniques demonstrate that stem-cell-fate specification is an extremely complex process, regulated by multiple mutually interacting molecular mechanisms involving multiple regulatory feedback loops. Given this complexity and the sensitive dependence of stem cell differentiation on signaling cues from the extracellular environment, how are we best to develop a coherent quantitative understanding of stem cell fate at the systems level? One approach that we and other researchers have begun to investigate is the application of techniques derived in the computational disciplines (mathematics, physics, computer science, etc.) to problems in stem cell biology. Here, we briefly sketch a few pertinent results from the literature in this area and discuss future potential applications of computational techniques to stem cell systems biology.

Mathematical modelling of skeletal repair.

Tissue engineering offers significant promise as a viable alternative to current clinical strategies for replacement of damaged tissue as a consequence of disease or trauma. Since mathematical modelling is a valuable tool in the analysis of complex systems, appropriate use of mathematical models has tremendous potential for advancing the understanding of the physical processes involved in such tissue reconstruction. In this review, the potential benefits, and limitations, of theoretical modelling in tissue engineering applications are examined with specific emphasis on tissue engineering of bone. A central tissue engineering approach is the in vivo implantation of a biomimetic scaffold seeded with an appropriate population of stem or progenitor cells. This review will therefore consider the theory behind a number of key factors affecting the success of such a strategy including: stem cell or progenitor population expansion and differentiation ex vivo; cell adhesion and migration, and the effective design of scaffolds; and delivery of nutrient to avascular structures. The focus will be on current work in this area, as well as on highlighting limitations and suggesting possible directions for future work to advance health-care for all.

Prediction of Bayley and Stanford-Binet scores with a group of very low birthweight children.

AIM: To study the prediction of cognitive development with a group of very low birthweight infants (<1500 g) at 18 months and at 4 years of age. METHODS: Bayley Scales of Infant Development-II Mental Development Indexes (MDI), and Stanford-Binet Intelligence Scale (S-B) Composite Scores were studied in a population of 334 children with birthweights <1500 g. Independent variables measured were gestational age, birthweight, gender and parental socio-economic status (SES). RESULTS: Longer gestation (28 weeks and over) and higher birthweight (1000-1500 g) proved to be advantageous for cognitive ability at both 18 months and 4 years. Other significant advantages were associated with female and higher SES. High correlations were found between MDI and the S-B Composite Score (r = 0.62), and between birthweight and gestation (r = 0.72). When information available at birth was included in forward stepwise regression analyses to predict the S-B Composite at 4 years, the best predictors were MDI at 18 months and SES. CONCLUSION: With the measures employed and this population, prediction of cognitive development from early childhood to preschool was possible. This may enable reliable identification of those children at risk for delayed cognitive development who require intervention before starting school.

The Lys1010-Lys1325 fragment of the Wilson's disease protein binds nucleotides and interacts with the N-terminal domain of this protein in a copper-dependent manner.

Wilson's disease, an autosomal disorder associated with vast accumulation of copper in tissues, is caused by mutations in a gene encoding a copper-transporting ATPase (Wilson's disease protein, WNDP). Numerous mutations have been identified throughout the WNDP sequence, particularly in the Lys(1010)-Lys(1325) segment; however, the biochemical properties and molecular mechanism of WNDP remain poorly characterized. Here, the Lys(1010)-Lys(1325) fragment of WNDP was overexpressed, purified, and shown to form an independently folded ATP-binding domain (ATP-BD). ATP-BD binds the fluorescent ATP analogue trinitrophenyl-ATP with high affinity, and ATP competes with trinitrophenyl-ATP for the binding site; ADP and AMP appear to bind to ATP-BD at the site separate from ATP. Purified ATP-BD hydrolyzes ATP and interacts specifically with the N-terminal copper-binding domain of WNDP (N-WNDP). Strikingly, copper binding to N-WNDP diminishes these interactions, suggesting that the copper-dependent change in domain-domain contact may represent the mechanism of WNDP regulation. In agreement with this hypothesis, N-WNDP induces conformational changes in ATP-BD as evidenced by the altered nucleotide binding properties of ATP-BD in the presence of N-WNDP. Significantly, the effects of copper-free and copper-bound N-WNDP on ATP-BD are not identical. The implications of these results for the WNDP function are discussed.

Some influences on cognitive development in a group of very low birthweight infants at four years.

AIM: To report the findings of an audit of 91 < 1,500 grams birthweight children at four years of age and consider the cognitive and behavioural results in the light of gestation, birthweight and socio-economic status. METHOD: Children born in 1993 were assessed using the Stanford-Binet Intelligence Scale, Conners' Parent Rating Scales and clinical evaluation. Parent interview provided further information on development, health and behaviour. RESULTS: Significant differences were demonstrated in a number of comparisons. For the Stanford-Binet results, children >28 weeks gestation obtained higher mean scores than those of shorter gestation for Abstract Visual Reasoning (representing cognitive skills such as spatial perception and visual - motor coordination), and Quantitative Reasoning (which includes numerical ability). In the case of birthweight, the mean score for children > or = 1,000 grams was higher on Quantitative Reasoning than that obtained by the lower birthweight group. When socioeconomic status classification was introduced, a number of differences in test scores were apparent, with children whose parents were in the lowest classification consistently performing less well. Quality of language results followed the same pattern. Analysis of the Conners' Scales identified this low socio-economic status group as having more learning problems and higher scores on the Hyperactivity Index than those whose parents were in the highest socioeconomic status group. CONCLUSION: Results for cognitive performance on the Stanford-Binet Scale were within one standard deviation of the mean, for the study population as a group. However, findings suggested that children <1,000 grams birthweight, or <28 weeks gestation, and those <1,500 grams with parents in the lowest socio-economic group, were at particular risk of achieving lower scores prior to school entry. Their pattern of cognitive ability and behaviour could inhibit their adjustment to formal education when complex concepts require problem solving at a more advanced level.

Developmental outcome at 18 months of children less than 1000 grams.

AIMS: Aims of this paper were to carry out an audit of 105 extremely low birth weight infants at 18 months of age, identifying problems, disseminating the resulting information and providing a basis for future work. METHODS: Children born in 1990-2 were classified in categories I to IV according to outcome, and selected perinatal variables were analysed for these groupings. RESULTS: The disability rate (categories I and II) within this cohort was 21% a similar finding to that reported in other literature. For the group with slow motor development and/or tonal abnormalities the percentage was 15. Despite the high risk nature of this group 64% of children were progressing well (category IV) at this age. No significant differences in outcome were found between small for gestational age and appropriate weight for gestational age infants. Significant results were demonstrated in the adverse effects of chronic lung disease and intraventricular haemorrhage on subsequent development. CONCLUSION: The information obtained from this study provides support for the use of this type of audit in Units with extremely low birth weight populations.

Cognitive development and behaviour in very low birthweight twins at four years.

This study included two groups of 37 children, one of twins and the other singletons at 4 years of age. All subjects has birthweights under 1500 grams and individuals in the groups were matched for birth date, gender and birthweight. Except when parental socio-economic status was taken into account, no significant differences between twins and singletons were observed on any of the results of The Stanford-Binet Intelligence Scale, nor were there any when the twins and singletons were divided into groups with birthweights < 1000 grams and 1000 to 1499 grams. When cognitive scores were analysed in relation to socio-economic status, there were significant differences in the whole population between subjects in the high and low socioeconomic status groups, with higher mean scores for the former. Comparison of the twins and singletons with parents in the lower socio-economic status group did not produce any significant differences but in the case of the upper socio-economic status group the singletons scored significantly better than the twins in Quantitative Reasoning and on the Composite Score. No significant differences were demonstrated in the clinical assessment of speech, language or behaviour. So far as general life considerations and health were concerned only one significant difference was found and this was for the number of siblings born subsequently, with more born in the singleton families. This study did not provide support for the view that singletons and twins differ significantly in the areas considered.

A semi-micro method for fibrinogen determination essentially unaffected by turbidity.

A semi-micro method (BR BLUE) is presented using Coomassie Brilliant Blue G-250 color reagent for the determination of fibrinogen on potentially turbid specimens. It is compared with the reference Clauss method using the automated MLA Electra 1000C Automatic Coagulation Timer and the semi-automated Mechrolab Clot Timer. Correlation is excellent (r = 0.97). The BR BLUE method is also seen to be reasonably precise and linear.

Perinatal care crucial for very premature babies' later development.

Thanks to improved medical technology, babies born very prematurely now have a much greater chance of surviving than they used to. A research team from the University of Auckland and National Women's Hospital is examining the long-term effects of perinatal difficulties on children weighing under 1500g at birth. One major aim is to discover whether the increasing number of immature infants who survive has led to a greater proportion of children with problems. The group's findings indicate that this is not so. Dr Barton MacArthur (Education), Associate-Professor Ross Howie (Paediatrics and National Women's Hospital) and Dr Anne Dezoete (MNational Women's Hospital, Child Development Unit) prepared this report on their cooperative research association.

Transient neonatal tyrosinaemia.

Children who had presented with transient neonatal tyrosinaemia (TNT) were compared with a group of unaffected controls at 7-9 years of age. A comprehensive psychometric assessment revealed significant differences between the groups in adaptive behaviour, psycholinguistic abilities, and speed of learning. In nearly all components of the tests used, higher levels of TNT were associated with lower levels of performance. This study demonstrates that TNT, a condition commonly regarded as benign in the short term, has long-term effects which may be detrimental to the child in school.

Comparison of infants dying from the sudden infant death syndrome with matched live controls.

Development, medical history and social background of 79 sudden infant death syndrome (SIDS) and 79 age matched controls were compared. SIDS was associated with a high proportion of S.G.A. infants, more hospital admissions and increased reporting of symptoms, particularly irritability, by parents. Social factors did not play a significant role with no evidence of marked social deprivation or lack of caring in either group. When this background data from both groups was compared there was no evidence of reliable predictors of sudden infant death.

Sudden infant death syndrome in south east Scotland.

Three hundred and fifty eight infants from south east Scotland who died suddenly were classified into four groups. Categories for these groups ranged from where a definite cause of death had been recorded to where no explanation had been provided and no associated disorder was discovered (SIDS). Our results supported the view that there are few differences in the history of cases certified as SIDS and other cases reported as dying suddenly but with an explanation. Groups that most closely matched the SIDS definition employed were reported to be healthier throughout life and freer from illness in the 48 hours before death. From the findings of this study the 'true' SIDS group did not appear as an 'at risk' population. The study group as a whole was not marked by social deprivation, poor mothering, or less privileged families. The importance of intensive investigation, including postmortem examination was emphasised, as misdiagnosis may give a 'falsely' inflated picture of the incidence of the syndrome and could cause unnecessary anxiety.

A study of the sudden infant death syndrome by age.

Data from 201 infants in South East Scotland certified at death as sudden infant death syndrome (SIDS) were analysed to determine whether the cases in the younger age groups at death (4-19 weeks) differed from those in the older group (19-52 weeks) on a number of background, developmental and medical variables frequently associated with SIDS in research findings. Some of the findings of the study were contrary to much of the research on SIDS in general. While some factors commonly associated with SIDS were found in both groups the 'typical' pattern presented in SIDS in the literature was more strongly associated with the older group. Marked differences between the two age groupings at death raise questions concerning those aspects of causation and aetiology that may go undetected if this distinction is not made during the analysis of SIDS information.

A case of agammaglobulinaemia complicated by meningoencephalitis due to Echo virus 27.

A boy with agammaglobulinaemia diagnosed and treated from the age of 3 years 2 months developed encephalitis at 4 years 3 months. The illness showed a remitting but deteriorating course until death aged 6 years 5 months. Echo virus type 27 was isolated from the cerebrospinal fluid during the terminal illness. The virus was not isolated from five earlier cerebrospinal fluid samples nor was virus detected in a temporal lobe biopsy.

School progress and cognitive development of 6-year-old children whose mothers were treated antenatally with betamethasone.

The cognitive development of children whose mothers had been included in the first Auckland trials of betamethasone therapy in premature labor were studied. An earlier study of these children used psychometric tests during the fifth year of life. In the present study tests were given during the seventh year of life (the second year of school) to 250 (82.2%) of 304 surviving children. Of the 250 children, 139 were in the group whose mothers had received betamethasone and 111 were in the control group. Further tests of cognitive development were made, together with assessment of the children's progress in school. Again, on the majority of measures there were no significant differences between children whose mothers had received betamethasone and the children in the control group. Calculations of statistical power showed that important differences were unlikely to have been missed.