Machine learning guided aptamer refinement and discovery.

Aptamers are single-stranded nucleic acid ligands that bind to target molecules with high affinity and specificity. They are typically discovered by searching large libraries for sequences with desirable binding properties. These libraries, however, are practically constrained to a fraction of the theoretical sequence space. Machine learning provides an opportunity to intelligently navigate this space to identify high-performing aptamers. Here, we propose an approach that employs particle display (PD) to partition a library of aptamers by affinity, and uses such data to train machine learning models to predict affinity in silico. Our model predicted high-affinity DNA aptamers from experimental candidates at a rate 11-fold higher than random perturbation and generated novel, high-affinity aptamers at a greater rate than observed by PD alone. Our approach also facilitated the design of truncated aptamers 70% shorter and with higher binding affinity (1.5 nM) than the best experimental candidate. This work demonstrates how combining machine learning and physical approaches can be used to expedite the discovery of better diagnostic and therapeutic agents.

Continued Transmission of HIV Among Young Adults Who Inject Drugs in San Francisco: Still Room for Improvement.

We measured HIV incidence rate, trend and risk factors in 564 HIV-negative young people (< 30 years) who inject drugs (PWID) in San Francisco between 2000 and 2014. HIV incidence was 0.93/100 person-years (PY; 95% CI 0.50, 1.73). Incidence varied between 0.62/100 PY in 2000-2002 and 1.06/100 PY in 2012-2014 (P for trend = 1.0). HIV incidence varied significantly (P < 0.01) by race/ethnicity: among Hispanics it was 8.19/100 PY (95% CI 3.41, 19.68), African-Americans 4.59/100 PY (95% CI 1.15, 18.37), and Whites 0.26/100 PY (95% CI 0.06, 1.03). Male participants who reported sex with men (MSM) had higher HIV incidence (2.63/100 PY; 95% CI 1.31, 5.25) compared to males who did not report MSM (0.50/100 PY; 95% CI 0.12, 1.99) (P = 0.01). Despite an overall stable HIV incidence trend, incidence was elevated among African-American and Hispanic PWID, and men who have sex with men. Addressing prevention needs in these key populations is critical for the goal of eliminating HIV transmission.

Evaluation of dried blood spot protocols with the Bio-Rad GS HIV Combo Ag/Ab EIA and Geenius™ HIV 1/2 Supplemental Assay.

OBJECTIVE: FDA-approved antigen/antibody combo and HIV-1/2 differentiation supplemental tests do not have claims for dried blood spot (DBS) use. We compared two DBS-modified protocols, the Bio-Rad GS HIV Combo Ag/Ab (BRC) EIA and Geenius™ HIV-1/2 (Geenius) Supplemental Assay, to plasma protocols and evaluated them in the CDC/APHL HIV diagnostic algorithm. METHODS: BRC-DBS p24 analytical sensitivity was calculated from serial dilutions of p24. DBS specimens included 11 HIV-1 seroconverters, 151 HIV-1-positive individuals, including 20 on antiretroviral therapy, 31 HIV-2-positive and one HIV-1/HIV-2-positive individuals. BRC-reactive specimens were tested with Geenius using the same DBS eluate. Matched plasma specimens were tested with BRC, an IgG/IgM immunoassay and Geenius. DBS and plasma results were compared using the McNemar's test. A DBS-algorithm applied to 348 DBS from high-risk individuals who participated in surveillance was compared to HIV status based on local testing algorithms. RESULTS: BRC-DBS detects p24 at a concentration 18 times higher than in plasma. In seroconverters, BRC-DBS detected more infections than the IgG/IgM immunoassay in plasma (p=0.0133), but fewer infections than BRC-plasma (p=0.0133). In addition, the BRC/Geenius-plasma algorithm identified more HIV-1 infections than the BRC/Geenius-DBS algorithm (p=0.0455). The DBS protocols correctly identified HIV status for established HIV-1 infections, including those on therapy, HIV-2 infections, and surveillance specimens. CONCLUSIONS: The DBS protocols exhibited promising performance and allowed rapid supplemental testing. Although the DBS algorithm missed some early infections, it showed similar results when applied to specimens from a high-risk population. Implementation of a DBS algorithm would benefit testing programs without capacity for venipuncture.

The collective impact of rare diseases in Western Australia: an estimate using a population-based cohort.

PURPOSE: It has been argued that rare diseases should be recognized as a public health priority. However, there is a shortage of epidemiological data describing the true burden of rare diseases. This study investigated hospital service use to provide a better understanding of the collective health and economic impacts of rare diseases. METHODS: Novel methodology was developed using a carefully constructed set of diagnostic codes, a selection of rare disease cohorts from hospital administrative data, and advanced data-linkage technologies. Outcomes included health-service use and hospital admission costs. RESULTS: In 2010, cohort members who were alive represented approximately 2.0% of the Western Australian population. The cohort accounted for 4.6% of people discharged from hospital and 9.9% of hospital discharges, and it had a greater average length of stay than the general population. The total cost of hospital discharges for the cohort represented 10.5% of 2010 state inpatient hospital costs. CONCLUSIONS: This population-based cohort study provides strong new evidence of a marked disparity between the proportion of the population with rare diseases and their combined health-system costs. The methodology will inform future rare-disease studies, and the evidence will guide government strategies for managing the service needs of people living with rare diseases.Genet Med advance online publication 22 September 2016.

Use of mechanical airway clearance devices in the home by people with neuromuscular disorders: effects on health service use and lifestyle benefits.

BACKGROUND: People with neuromuscular disorders (NMD) exhibit weak coughs and are susceptible to recurrent chest infections and acute respiratory complications, the most frequent reasons for their unplanned hospital admissions. Mechanical insufflation-exsufflation (MI-E) devices are a non-invasive method of increasing peak cough flow, improving cough efficacy, the clearance of secretion and overcoming atelectasis. There is limited published evidence on the impact of home use MI-E devices on health service utilisation. The aims of the study were: to assess the self-reported health and lifestyle benefits experienced as a result of home use of MI-E devices; and evaluate the effects of in-home use of MI-E devices on Emergency Department (ED) presentations, hospital admissions and inpatient length of stay (LOS). METHODS: Individuals with NMD who were accessing a home MI-E device provided through Muscular Dystrophy Western Australia were invited to participate in a quantitative survey to obtain information on their experiences and self-assessed changes in respiratory health. An ad-hoc record linkage was performed to extract hospital, ED and mortality data from the Western Australian Department of Health (DOHWA). The main outcome measures were ED presentations, hospital separations and LOS, before and after commencement of home use of an MI-E device. RESULTS: Thirty seven individuals with NMD using a MI-E device at home consented to participate in this study. The majority (73%) of participants reported using the MI-E device daily or weekly at home without medical assistance and 32% had used the machine to resolve a choking episode. The survey highlighted benefits to respiratory function maintenance and the ability to manage increased health care needs at home. Not using a home MI-E device was associated with an increased risk of ED presentations (RR = 1.76, 95% CI 1.1-2.84). The number of hospital separations and LOS reduced after the use of MI-E device, but not significantly. No deaths were observed in participants using the MI-E device at home. CONCLUSIONS: Home use of a MI-E device by people living with NMD may have a potential impact on reducing their health service utilisation and risk of death. Future research with greater subject numbers and longer follow-up periods is recommended to enhance this field of study.

The custodian administered research extract server: "improving the pipeline" in linked data delivery systems.

BACKGROUND: At Western Australia's Data Linkage Branch (DLB) the extraction of linked data has become increasingly complex over the past decade and classical methods of data delivery are unsuited to the larger extractions which have become the norm. The Custodian Administered Research Extract Server (CARES) is a fast, accurate and predictable approach to linked data extraction. METHODS: The Data Linkage Branch (DLB) creates linkage keys within and between datasets. To comply with the separation principal, these keys are sent to applicable data collection agencies for extraction. Routing requests through multiple channels is inefficient and makes it hard to monitor work and predict delivery times. CARES was developed to address these shortcomings and involved ongoing consultation with the Custodians and staff of collections, plus challenges of hardware, programming, governance and security. RESULTS: The introduction of CARES has reduced the workload burden of linked data extractions, while improving the efficiency, stability and predictability of turnaround times. CONCLUSIONS: As the scope of a linkage system broadens, challenges in data delivery are inevitable. CARES overcomes multiple obstacles with no sacrifice to the integrity, confidentiality or security of data. CARES is a valuable component of linkage infrastructure that is operable at any scale and adaptable to many data environments.

Morbidity associated with amphetamine-related presentations to an emergency department: A record linkage study.

OBJECTIVE: Amphetamine use is a global public health problem. We examined hospitalisations in a cohort of 138 patients who presented with an amphetamine-related problem to an ED in 2005. METHODS: A record linkage study, using the morbidity, ED and mortality databases in the Data Linkage Unit of the Department of Health, Western Australia. The main outcome measures were hospital separations and length of stay (LOS) 5 years before and 4 years after entry into the cohort. RESULTS: One hundred and thirty patients (94%) with an amphetamine-related presentation had a link with the hospital morbidity dataset. The most common diagnosis before and after cohort entry was mental disorders (before: F00-F99; 405 separations, total LOS 2570 days; after: 309 separations, total LOS 3671 days). Injury and poisoning was the next most common in both time periods. Men had an increased relative risk (RR) for all days of psychiatric care (RR 2.12, 95% CI 1.04-4.35). After adjusting for age and sex, the highest risks of increased LOS occurred within 1 year before (RR 2.22, 95% CI 1.01-4.91) and 2 years post entry into the cohort (RR 4.21, 95% CI 1.87-9.46 and RR 2.82, 95% CI 1.25-6.34). There were four (2.9%, 95% CI 0.9-7.7%) deaths, which occurred within 2 years post cohort entry. CONCLUSION: Amphetamine-related presentations to the ED are associated with a significant cluster of hospitalisations around that episode. This is most prominent for psychiatric diagnoses, with a large increase in the total LOS in the year following cohort entry. Counselling less risky behaviour might decrease the burden of illness.

Adjusting for under-identification of Aboriginal and/or Torres Strait Islander births in time series produced from birth records: using record linkage of survey data and administrative data sources.

BACKGROUND: Statistical time series derived from administrative data sets form key indicators in measuring progress in addressing disadvantage in Aboriginal and Torres Strait Islander populations in Australia. However, inconsistencies in the reporting of Indigenous status can cause difficulties in producing reliable indicators. External data sources, such as survey data, provide a means of assessing the consistency of administrative data and may be used to adjust statistics based on administrative data sources. METHODS: We used record linkage between a large-scale survey (the Western Australian Aboriginal Child Health Survey), and two administrative data sources (the Western Australia (WA) Register of Births and the WA Midwives' Notification System) to compare the degree of consistency in determining Indigenous status of children between the two sources. We then used a logistic regression model predicting probability of consistency between the two sources to estimate the probability of each record on the two administrative data sources being identified as being of Aboriginal and/or Torres Strait Islander origin in a survey. By summing these probabilities we produced model-adjusted time series of neonatal outcomes for Aboriginal and/or Torres Strait Islander births. RESULTS: Compared to survey data, information based only on the two administrative data sources identified substantially fewer Aboriginal and/or Torres Strait Islander births. However, these births were not randomly distributed. Births of children identified as being of Aboriginal and/or Torres Strait Islander origin in the survey only were more likely to be living in urban areas, in less disadvantaged areas, and to have only one parent who identifies as being of Aboriginal and/or Torres Strait Islander origin, particularly the father. They were also more likely to have better health and wellbeing outcomes. Applying an adjustment model based on the linked survey data increased the estimated number of Aboriginal and/or Torres Strait Islander births in WA by around 25%, however this increase was accompanied by lower overall proportions of low birth weight and low gestational age babies. CONCLUSIONS: Record linkage of survey data to administrative data sets is useful to validate the quality of recording of demographic information in administrative data sources, and such information can be used to adjust for differential identification in administrative data.

Morbidity associated with heroin overdose presentations to an emergency department: a 10-year record linkage study.

INTRODUCTION: To examine hospitalizations in a cohort of 224 patients who presented with non-fatal heroin overdose to an ED. METHODS: A record linkage study, using the morbidity, mental health and mortality databases in the Data Linkage Unit of the Department of Health, Western Australia. The main outcome measures were hospital separations 5 years before and after entry into the cohort. RESULTS: Before entry into the cohort, 199 (89%) patients had an admission to mental health services. These 199 had a combined total of 1367 separations, most commonly for a mental health condition, injury or poisoning. Women had more than twice the relative risk (RR) of men for all separations (RR 2.35, 95% confidence interval [CI] 1.96-2.82, P < 0.001) and for injury and poisoning separations (RR 2.04, 95% CI 1.56-2.66, P < 0.001). The highest concentrations of separations occurred within 1 year before and 1 year after entry into the cohort. There were 12 (5.4%, 95% CI 2.9-9.4%) deaths, most commonly from overdose. CONCLUSION: Non-fatal heroin overdose ED presentations are associated with a cluster of hospitalizations around that episode, likely to be related to heroin availability. Presentation to hospital by heroin users represents an opportunity to counsel less risky behaviour.