A Boolean-based machine learning framework identifies predictive biomarkers of HSP90-targeted therapy response in prostate cancer

Precision medicine has emerged as an important paradigm in oncology, driven by the significant heterogeneity of individual patients' tumour. A key prerequisite for effective implementation of precision oncology is the development of companion biomarkers that can predict response to anti-cancer therapies and guide patient selection for clinical trials and/or treatment. However, reliable predictive biomarkers are currently lacking for many anti-cancer therapies, hampering their clinical application. Here, we developed a novel machine learning-based framework to derive predictive multi-gene biomarker panels and associated expression signatures that accurately predict cancer drug sensitivity. We demonstrated the power of the approach by applying it to identify response biomarker panels for an Hsp90-based therapy in prostate cancer, using proteomic data profiled from prostate cancer patient-derived explants. Our approach employs a rational feature section strategy to maximise model performance, and innovatively utilizes Boolean algebra methods to derive specific expression signatures of the marker proteins. Given suitable data for model training, the approach is also applicable to other cancer drug agents in different tumour settings.

The expected burden of mesothelioma mortality in Great Britain from 2002 to 2050

The British mesothelioma register contains all deaths from 1968 to 2001 where mesothelioma was mentioned on the death certificate. These data were used to predict the future burden of mesothelioma mortality in Great Britain. Poisson regression analysis was used to model male mesothelioma deaths from 1968 to 2001 as a function of the rise and fall of asbestos exposure during the 20th century, and hence to predict numbers of male deaths in the years 2002–2050. The annual number of mesothelioma deaths in Great Britain has risen increasingly rapidly from 153 deaths in 1968 to 1848 in 2001 and, using our preferred model, is predicted to peak at around 1950 to 2450 deaths per year between 2011 and 2015. Following this peak, the number of deaths is expected to decline rapidly. The eventual death rate will depend on the background level and any residual asbestos exposure. Between 1968 and 2050, there will have been approximately 90 000 deaths from mesothelioma in Great Britain, 65 000 of which will occur after 2001.

Comparison of Health and Safety Executive and Cumbrian birth cohort studies of risk of leukaemia/non-Hodgkin's lymphoma in relation to paternal preconceptional irradiation

In 1993, a case-control study by the Health and Safety Executive (HSE) assessed the risk of leukaemia and non-Hodgkin's lymphoma (LNHL) among children of fathers employed at the Sellafield nuclear installation in relation to paternal preconceptional irradiation (PPI). It concluded that the statistical association between risk of LNHL and PPI was confined to children born in the village of Seascale, where the dose-response was extremely high and very significant. In contrast, in 2002, a Cumbrian birth cohort study, investigating largely the same cases, concluded that this statistical association was not significantly different among children born inside and outside Seascale and estimated the dose-response inside Seascale to be much lower. This review makes a detailed comparison of the two studies, considering their design, data and analyses. The differences between their findings are due to: (i) differences in the distribution of offspring-years which are differential with respect to dose category and Seascale birth status, (ii) a non-Seascale high-dose case included in the Cumbrian but not the HSE study, (iii) differences between analyses using categorical and continuous PPI dose and (iv) the presence of Seascale controls with PPI over 200 mSv in the Cumbrian but not the HSE study.

Investigating and analysing workplace clusters of diseases: a Health & Safety Executive perspective

BACKGROUND

Clusters of disease arising in workplaces cause concern among the management of the company, the workers affected and their families and friends. Chance is the most likely explanation for their occurrence, although a number of real workplace hazards have been identified through their observation and investigation. Employers have a duty to investigate such occurrences in order to assess whether some unknown or unidentified hazard is at work and to take the appropriate action. Several papers have been published over the last 15 years or so that set out a method for investigating workplace clusters of disease. Aims This paper presents the steps in the approach taken by the Health & Safety Executive in Great Britain.

METHOD

An initial step identifies the relevant stakeholders at the outset, in order to maintain a realistic expectation of what the investigation can hope to achieve and to open a dialogue. The main steps in the assessment are: (1) identifying cases; (2) determining the other parameters of the investigation; (3) statistically assessing the cluster; (4) examining potential exposures and assessing their biological plausibility; and (5) determining the overall significance of the cluster. The approach is illustrated throughout by examples.

The quantitative risks of mesothelioma and lung cancer in relation to asbestos exposure

Mortality reports on asbestos exposed cohorts which gave information on exposure levels from which (as a minimum) a cohort average cumulative exposure could be estimated were reviewed. At exposure levels seen in occupational cohorts it is concluded that the exposure specific risk of mesothelioma from the three principal commercial asbestos types is broadly in the ratio 1:100:500 for chrysotile, amosite and crocidolite respectively. For lung cancer the conclusions are less clear cut. Cohorts exposed only to crocidolite or amosite record similar exposure specific risk levels (around 5% excess lung cancer per f/ml.yr); but chrysotile exposed cohorts show a less consistent picture, with a clear discrepancy between the mortality experience of a cohort of xhrysotile textile workers in Carolina and the Quebec miners cohort. Taking account of the excess risk recorded by cohorts with mixed fibre exposures (generally<1%), the Carolina experience looks uptypically high. It is suggested that a best estimate lung cancer risk for chrysotile alone would be 0.1%, with a highest reasonable estimate of 0.5%. The risk differential between chrysotile and the two amphibole fibres for lunc cancer is thus between 1:10 and 1:50. Examination of the inter-study dose response relationship for the amphibole fibres suggests a non-linear relationship for all three cancer endpoints (pleural and peritoneal mesotheliomas, and lung cancer). The peritoneal mesothelioma risk is proportional to the square of cumulative exposure, lung cancer risk lies between a linear and square relationship and pleural mesothelioma seems to rise less than linearly with cumulative dose. Although these non-linear relationships provide a best fit ot the data, statistical and other uncertainties mean that a linear relationship remains arguable for pleural and lung tumours (but not or peritoneal tumours). Based on these considerations, and a discussion fo the associated uncertainties, a series of quantified risk summary statements for different elvels of cumulative exposure are presented.

Spontaneous abortion in the British semiconductor industry: An HSE investigation. Health and Safety Executive

BACKGROUND

The UK Health and Safety Executive (HSE) conducted a study to examine the risk of spontaneous abortion (SAB) in British female semiconductor industry workers, following reports from the USA which suggested an association between risk of SAB and work in fabrication rooms and/or exposure to ethylene glycol ethers.

METHODS

A nested case-control study based on 2,207 women who had worked at eight manufacturing sites during a 5-year retrospective time frame was established; 36 cases were matched with 80 controls.

RESULTS

The overall SAB rate in the industry was 10.0%. (65 SABs/651 pregnancies) The crude odds ratio (OR) for fabrication work was 0.65 (95% CI 0.30-1.40). This was essentially unchanged after adjustment for a range of potential confounding factors in the first 3 months of pregnancy and was reduced to 0.58 (95% CI 0.26-1.30) after adjustment for smoking in the previous 12 months. There were no statistically significantly elevated ORs for any work group or any specific chemical or physical exposure in the industry.

CONCLUSIONS

There is no evidence of an increased risk of SAB in the British semiconductor industry. Am. J. Ind. Med. 36:557-572, 1999. Published 1999 Wiley-Liss, Inc.

Maternal occupational exposure to electromagnetic fields before, during, and after pregnancy in relation to risks of childhood cancers: findings from the Oxford Survey of Childhood Cancers, 1953-1981 deaths

BACKGROUND

The concern that maternal exposure to electromagnetic fields (EMF) might be related to childhood cancer risks, particularly leukemia risks.

METHODS

Maternal occupational data already collected as part of the Oxford Survey of Childhood Cancers have been reviewed. Information on occupations held before, during, and after the pregnancy was sought for 15,041 children dying of cancer in Great Britain in the period 1953-1981, and for an equal number of matched controls. Each period of working was classified under one of five headings: (1) sewing machinist; (2) textile industry workers (other than sewing machinists) with likely exposures to EMF; (3) other machinists and other jobs with likely "higher" EMF exposure; (4) other jobs with likely exposure to some EMF, and (5) jobs with little potential for EMF exposure.

RESULTS

Relative to risks in the children of mothers who held occupations with little potential for EMF exposure during pregnancy (a category that included housewives), risks of all childhood cancers were close to unity both for the children of sewing machinists (22 case and 31 control mothers, RR 0.72, 95% CI 0.42 to 1.25) and for the children of other machinists with likely "higher" EMF exposures (44 case and 47 control mothers, RR 0.93, 95% CI 0.61 to 1.41). Corresponding risks for all childhood leukemias and for all childhood brain cancers were similarly unexceptional. Simultaneous adjustment for social class, maternal age at birth of child, and sibship position had little effect.

CONCLUSIONS

The study findings did not indicate that maternal occupational exposure to EMF during pregnancy is a risk factor for childhood leukemias, childhood brain cancers, or the generality of all childhood cancers.

Continuing increase in mesothelioma mortality in Britain

Mesothelioma is closely related to exposure to asbestos, and mesothelioma mortality can be taken as an index of past exposure to asbestos in the population. We analysed mesothelioma mortality since 1968 to assess the current state of the mesothelioma epidemic, and to predict its future course. We found that rates of mesothelioma in men formed a clear pattern defined by age and date of birth. Rates rose steeply with age showing a very similar pattern in all five-year birth cohorts. By date of birth, rates increased from mid-1893 to mid-1948, and then fell. Relative to the 1943-48 cohort, the risk for the 1948-53 cohort is 0.79 and for the 1953-58 cohort 0.48. Despite these falls, if the age profile of rates for these cohorts follows the pattern of past cohorts, their predicted lifetime mesothelioma risks will be 1.3%, 1.0%, and 0.6%. Combining projections for all cohorts results in a peak of annual male mesothelioma deaths in about the year 2020 of between 2700 and 3300 deaths. If diagnostic trend is responsible for a 20% growth in recorded cases every 5 years--an extreme but arguable case--and if this trend has now ceased, the peak of annual male deaths will be reduced to 1300, reached around the year 2010. Analysis of occupations recorded on death certificates indicate that building workers, especially plumbers and gas fitters, carpenters and electricians are the largest high-risk group. These data indicate that mesothelioma deaths will continue to increase for at least 15 and more likely 25 years. For the worst affected cohorts--men born in the 1940s--mesothelioma may account for around 1% of all deaths. Asbestos exposure at work in construction and building maintenance will account for a large proportion of these deaths, and it is important that such workers should be aware of the risks and take appropriate precautions.

Mortality of a cohort of tin miners 1941-86

The mortality patterns of United Kingdom tin miners were examined in relation to calendar period and duration of underground work with particular attention to lung cancer and exposure to radon. Subjects were all men who had worked for at least one year between 1941 and 1984 at one of two United Kingdom tin mines and for whom a complete work history could be constructed from mine records. Standardised mortality ratios (SMRs) were calculated using national (England and Wales) rates. The pattern of SMRs in relation to potential explanatory variables was analysed using Poisson regression methods. Mortalities from lung cancer and silicosis (including silicotuberculosis) were significantly raised and showed a significant relation with duration of underground work (mortality from stomach cancer was raised in both underground and surface workers, but not significantly). Excess mortality from silica related disease declined steeply from 35% among workers first exposed before 1920 to 1% among those first exposed after 1950. Thirteen surface workers with known exposure to arsenic had high rates of lung and stomach cancer. The SMR for lung cancer showed a consistent pattern in relation to duration of underground exposure, rising from 83 (observed/expected = 8/9.6) for surface workers (without exposure to arsenic) to 447 (15/3.4) for workers with more than 30 years underground exposure. Examination of the SMR for lung cancer by total underground exposure, age, and time since last exposure gave rise to a model for the expression of risk which depends only on total exposure and time since exposure. The fitted model implies that the effect of exposure to radon in a given year has no effect on risk for 10 years, then rapidly rises to a maximum from which the excess risk then declines, halving every 4.3 years. There were no direct measurements of historic radon levels. A conservative estimate based on measurements taken since 1969 by the National Radiological Protection Board and the Mines and Quarries Inspectorate is that the annual dose to an underground worker was about 10 working level months (WLM). Given this assumption, the risk/exposure slope implied by the present data, and the model fitted to it, was somewhat lower than that given in the fourth Committee on the Biological Effects of Ionisation Radiation (BEIR IV) report (about 40% lower for lifetime exposures). The present data also imply different risks depending on the age at exposure, with relatively higher lifetime risks for exposure at older ages, and relatively lower risks for exposures at younger ages. In conclusion, there was a clear relation between exposure to radon and death from lung cancer. The relative risk of lung cancer due to exposure to radon was not constant in cessation of exposure. The lifetime excess risk of lung cancer implied by these data for 40 years exposure at the current statutory limit of four WLM a year starting at age 20, was about 8% (79 excess deaths per 1000 exposed), assuming average smoking habits among the exposed workers. Control of dust concentrations in the mines has substantially reduced--and may have eliminated--direct mortality from silica related disease.

Mortality of workers in the British cotton industry in 1968-1984

The mortality of 3458 cotton industry workers originally enrolled in a study of respiratory symptoms in the period 1968-1970 was followed to the end of 1984. Both the total mortality and the mortality from respiratory disease were less than expected, and they both decreased as length of service increased. However, for the subjects who initially reported byssinotic symptoms, the mortality from respiratory disease was slightly raised overall, and it increased with length of service. These patterns of mortality indicate a survivor effect (ie, a tendency for those with respiratory weakness to leave the industry), together with a long-term effect reflected in respiratory mortality on the health of those workers susceptible to the effects of cotton dust. The mortality from lung cancer was lower than expected, and it decreased with length of service. This finding is consistent with other observations that exposure to cotton dust may reduce the risk of lung cancer.

Guinea pig maximization test for skin sensitisation: the use of fewer test animals

Current European Community (Annex V) guidelines recommend the use of 20 test animals in the guinea pig maximisation test for skin sensitization. The suitability, for classification and labelling purposes, of reducing the number of test animals has been examined by analysing the results of 40 studies submitted to the Health and Safety Executive, and by the use of a mathematical model. Our results suggest that in most cases an experiment with ten test animals can be used to determine satisfactorily whether a substance should be labelled with the risk phrase "may cause sensitisation by skin contact". However, serious consideration should be given to the need for additional investigation if two or three of the ten test animals show a sensitisation response. The highest non-irritant concentration of a substance should be used at challenge. Clearer guidance in Annex V on evaluating challenge responses would be beneficial.

Mortality of asbestos workers in England and Wales 1971-81

A national study of British asbestos workers is briefly described and the mortality experience of 31 150 male asbestos workers in England and Wales who had been medically examined at least once as part of that survey is presented. The survey population is divided into workers with occupational exposure to asbestos before the inception of the 1969 Asbestos Regulations and those who worked with asbestos only after 1969. Of the 1128 who had died, 897 had worked before 1969; 34 of the death certificates received for these men mentioned mesothelioma and for another nine asbestosis was reported in the absence of mesothelioma or lung cancer. A statistically significant excess of lung cancer (SMR 136) was found. For the post-1969 workers, one case of asbestosis and one case of mesothelioma were reported, but further investigation of these cases showed probable occupational exposure to asbestos many years before 1969. The time from first exposure for this section of the population is too short to exclude an excess of asbestos related disease. The most noticeable excess of asbestos related disease was seen among the insulation workers who had more than twice (SMR 256) the expected number of deaths from lung cancer, and for whom almost 10% of all death certificates mentioned mesothelioma. No excess of any alimentary tract cancer was found and the population showed a significant deficit of large bowel cancer mortality (SMR 54).

Mortality of styrene production, polymerization and processing workers at a site in northwest England

The mortality of 622 men who worked for at least one year in the production, polymerization, and processing of styrene at a chemical site in the United Kingdom during the period 1945-1974 was surveyed up to the end of 1978; 3 072 male manual workers at the same site but unexposed to styrene were also studied. A statistically significant excess of lymphoma deaths was found in the exposed population, and two of the three deaths observed occurred in men less than 40 years of age. The small number of deaths and the lack of any evidence for an association with duration or level of exposure to styrene are reasons for interpreting this result cautiously. However, the otherwise normal pattern of mortality in the exposed population and the absence of any excess of lymphomas in the reference group lend some support to suggestions that exposure to styrene may be associated with lymphomas in man. An analysis of cancer registrations for the exposed population revealed no further cases of lymphoma but identified one case of lymphatic leukemia. An excess of laryngeal cancer registrations was found. As this effect has not been previously postulated and because of the small number of registrations, too much weight should not be attached to this one observation.

A mortality study of carbon black workers employed at five United Kingdom factories between 1947 and 1980

Mortality in 1422 male carbon black process workers with at least 12 months exposure was recorded from 1947 to 1980. Excess deaths from lung cancer, which were not statistically significant, were observed but interpretation is complicated by the incompleteness of data on the populations from two of the five factories studied. The highest excesses of lung cancer were in the two factories with incomplete data, which also had the lowest measured dust levels (though these were high). Furthermore, the duration of employment of lung cancer decedents was slightly less than for individually matched internal controls. Excess lung cancer after the tenth anniversaries of first exposure was 10 observed, 5.1 expected for the two factories with incomplete data; and 11 observed, 7.9 expected, for the other three factories.