Contraceptive-specific antimüllerian hormone values in reproductive-age women: a population study of 42,684 women.

OBJECTIVE: To determine how the contraceptive-specific serum antimüllerian hormone (AMH) levels compare across ages and percentiles in a reproductive-age cohort. DESIGN: Cross-sectional analysis of a prospectively recruited cohort. SETTING: Community. PATIENT(S): This study included US-based women of reproductive age who purchased a fertility hormone test and consented to participate in research between May 2018 and November 2021. At the time of hormone testing, participants were users of various contraceptives (combined oral contraceptive [n = 6,850], progestin-only pill [n = 465], hormonal [n = 4,867] or copper [n = 1,268] intrauterine device, implant [n = 834], vaginal ring [n = 886]) or women with regular menstrual cycles (n = 27,514). INTERVENTION(S): Contraceptive use. MAIN OUTCOME MEASURE(S): Age and contraceptive-specific estimates of AMH. RESULT(S): There were contraceptive-specific effects on AMH with effect estimates ranging from 0.83 (95% confidence interval [CI], 0.82-0.85) (17% lower) for the combined oral contraceptive pill to no effect (1.00; 95% CI, 0.98-1.03) for the hormonal intrauterine device. We did not observe age-specific differences in suppression. However, there were differential suppressive effects of the contraceptive method across AMH percentiles, with the greatest effect at lower percentiles and least effect at higher percentiles. For example, for women taking the combined oral contraceptive pill, the AMH level was 32% lower at the 10th percentile (coefficient, 0.68; 95% CI, 0.65-0.71), 19% lower at the 50th percentile (coefficient, 0.81; 95% CI, 0.79-0.84), and 5% lower at the 90th percentile (coefficient, 0.95; 95% CI, 0.92-0.98), with other forms of contraception showing similar discordances. CONCLUSION(S): These findings reinforce the body of literature that shows that hormonal contraceptives have different impacts on the AMH levels at a population level. These results add to this literature that these effects are not consistent; instead, the greatest impact occurs at the lower AMH percentiles. However, these contraceptive-dependent differences are small compared with the known biological variability in ovarian reserve at any given age. These reference values enable robust assessment of an individual's ovarian reserve relative to their peers without requiring cessation or potentially invasive removal of contraception.

Ethnic discordance in serum anti-Müllerian hormone in European and Indian healthy women and Indian infertile women.

RESEARCH QUESTION: Does anti-Müllerian hormone (AMH) differ between healthy European and Indian women, and are potential ethnic differences modified by infertility diagnosis? DESIGN: Cross-sectional analysis of three prospectively recruited cohorts (n = 2758); healthy European women (n = 758), healthy community cohort from Kolhapur, India (n = 400) and infertility cohort from Kolhapur, India (n = 1600). AMH was determined by assay. Ethnicity, age and cause of infertility were modelled using additive quantile regression models. RESULTS: Healthy Indian women had lower AMH than their healthy European counterparts (population estimates 20.0% lower [95% CI 7.2-36.5]), with increasing discordance with increasing age; at 25 years AMH was 11.9% lower (95% CI 9.4-14.1), increasing to 40.0% lower (95% CI 0-64.6) by age 45. Comparison of healthy and infertile Indian women revealed differences that were related to cause of infertility. Women whose male partner had severe oligoasthenoteratozoospermia (n = 95) had similar AMH to controls; women with polycystic ovary syndrome (n = 220) had higher AMH, especially in those <30 years, and in women with a principal diagnosis of unexplained infertility (n = 757) AMH was lower (median difference 22.6% lower; 95% CI 9.1-37.7) than controls. CONCLUSIONS: AMH is substantially lower in healthy Indian women at all ages than their European counterparts. Infertile Indian women have variable differences in AMH from healthy Indian controls, with the extent and direction of differences primarily reflecting the underlying cause of infertility. Recognition of ethnic and cause-specific differences are critical to ensure accurate contextualizing of results and clinical outcomes for patients.

Association of Epidural Analgesia in Women in Labor With Neonatal and Childhood Outcomes in a Population Cohort.

Importance: Although use of epidural analgesia during labor is safe, detailed information about its association with neonatal and child outcomes is limited. Objective: To investigate the association of labor epidural analgesia with neonatal outcomes and childhood development during the first 1000 days of life. Design, Setting, and Participants: This population-based cohort study used Scottish National Health Service hospital administrative data of all 435 281 singleton live births in Scotland between January 1, 2007, and December 31, 2016, with follow-up over the first 1000 days of life. All 435 281 mother-infant pairs delivering between 24 weeks 0 days and 43 weeks 6 days' gestation who were in active labor with cephalic presentation and who delivered vaginally or via unplanned cesarean delivery were included. Stillbirths and infants with known congenital anomalies were excluded. Data were analyzed between August 1, 2020, and July 23, 2021. Exposures: Epidural analgesia in labor. Main Outcomes and Measures: Neonatal outcomes included resuscitation, Apgar score less than 7 at 5 minutes, and neonatal unit admission. Childhood development measures (gross and fine motor function, communication, and social functioning) were obtained from standardized national childhood surveillance assessments performed at 2 years. Results: This study included a total of 435 281 live births with cephalic presentation in labor (median gestational age at delivery, 40 weeks [IQR, 39-41 weeks]; 221 153 male infants [50.8%]), of which 94 323 (21.7%) had labor epidural. Epidural analgesia was associated with a reduction in spontaneous vaginal deliveries (confounder-adjusted [Cadj] relative risk [RR], 0.46; 95% CI, 0.42-0.50), an increased risk of neonatal resuscitation (Cadj RR, 1.07; 95% CI, 1.03-1.11), and an increased risk of neonatal unit admission (Cadj RR, 1.14; 95% CI, 1.11-1.17). With additional analysis for mediation by mode of delivery (CMadj), these associations were reversed (CMadj RR, 0.83; 95% CI, 0.79-0.86 for neonatal resuscitation and CMadj RR, 0.94; 95% CI, 0.91-0.97 for neonatal unit admission). Epidural analgesia was associated with a reduced risk of an Apgar score less than 7 at 5 minutes in both confounder and confounder/mediation analyses. Epidural analgesia was associated with a reduced risk of having developmental concern in any domain at 2 years in confounder and confounder/mediation analyses (CMadj RR, 0.96; 95% CI, 0.93-0.98), with specifically fewer concerns regarding communication (CMadj RR, 0.96; 95% CI, 0.93-0.99) and fine motor skills (CMadj RR, 0.89; 95% CI, 0.82-0.97). Conclusions and Relevance: The results of this cohort study suggest that labor epidural analgesia is not independently associated with adverse neonatal or childhood development outcomes. Associations with neonatal resuscitation and admission were likely mediated by mode of delivery.

Identifying molecules as biosignatures with assembly theory and mass spectrometry.

The search for alien life is hard because we do not know what signatures are unique to life. We show why complex molecules found in high abundance are universal biosignatures and demonstrate the first intrinsic experimentally tractable measure of molecular complexity, called the molecular assembly index (MA). To do this we calculate the complexity of several million molecules and validate that their complexity can be experimentally determined by mass spectrometry. This approach allows us to identify molecular biosignatures from a set of diverse samples from around the world, outer space, and the laboratory, demonstrating it is possible to build a life detection experiment based on MA that could be deployed to extraterrestrial locations, and used as a complexity scale to quantify constraints needed to direct prebiotically plausible processes in the laboratory. Such an approach is vital for finding life elsewhere in the universe or creating de-novo life in the lab.

2008 financial crisis versus 2020 economic fallout: how COVID-19 might influence fertility treatment and live births.

RESEARCH QUESTION: The economic and reproductive medicine response to the coronavirus disease 2019 (COVID-19) pandemic in the USA has reduced the affordability and accessibility of fertility care. What is the impact of the 2008 financial recession and the COVID-19 recession on fertility treatments and cumulative live births? DESIGN: The study examined annual US natality, Centers for Disease Control and Prevention IVF cycle activity and live birth data from 1999 to 2018 encompassing 3,286,349 treatment cycles, to estimate the age-stratified reduction in IVF cycles undertaken after the 2008 financial recession, with forward quantitative modelling of IVF cycle activity and cumulative live births for 2020 to 2023. RESULTS: The financial recession of 2008 caused a 4-year plateau in fertility treatments with a predicted 53,026 (95% confidence interval [CI] 49,581 to 56,471) fewer IVF cycles and 16,872 (95% CI 16,713 to 17,031) fewer live births. A similar scale of economic recession would cause 67,386 (95% CI 61,686 to 73,086) fewer IVF cycles between 2020 and 2023, with women younger than 35 years overall undertaking 22,504 (95% CI 14,320 to 30,690) fewer cycles, compared with 4445 (95% CI 3144 to 5749) fewer cycles in women over the age of 40 years. This equates to overall 25,143 (95% CI 22,408 to 27,877) fewer predicted live births from IVF, of which only 490 (95% CI 381 to 601) are anticipated to occur in women over the age of 40 years. CONCLUSIONS: The COVID-19 recession could have a profound impact on US IVF live birth rates in young women, further aggravating pre-existing declines in total fertility rates.

Neonatal and early childhood outcomes following maternal anesthesia for cesarean section: a population-based cohort study.

BACKGROUND: The fetus is vulnerable to maternal drug exposure. We determined associations of exposure to spinal, epidural, or general anesthesia on neonatal and childhood development outcomes during the first 1000 days of life. METHODS: Population-based study of all singleton, cesarean livebirths of 24+0 to 43+6 weeks gestation between January 2007 and December 2016 in Scotland, stratified by urgency with follow-up to age 2 years. Models were adjusted for: maternal age, weight, ethnicity, socioeconomic status, smoking, drug-use, induction, parity, previous cesarean or abortion, pre-eclampsia, gestation, birth weight, and sex. RESULTS: 140 866 mothers underwent cesarean section (41.2% (57,971/140,866) elective, 58.8% (82,895/140,866) emergency) with general anesthesia used in 3.2% (1877/57,971) elective and 9.8% (8158/82,895) of emergency cases. In elective cases, general anesthesia versus spinal was associated with: neonatal resuscitation (crude event rate 16.2% vs 1.9% (adjusted RR 8.20, 95% CI 7.20 to 9.33), Apgar <7 at 5 min (4.6% vs 0.4% (adjRR 11.44, 95% CI 8.88 to 14.75)), and neonatal admission (8.6% vs 4.9% (adjRR 1.65, 95% CI 1.40 to 1.94)). Associations were similar in emergencies; resuscitation (32.2% vs 12.3% (adjRR 2.40, 95% CI 2.30 to 2.50)), Apgar <7 (12.6% vs 2.8% (adjRR 3.87, 95% CI 3.56 to 4.20), and admission (31.6% vs 19.9% (adjRR 1.20, 95% CI 1.15 to 1.25). There was a weak association between general anesthesia in emergency cases and having ≥1 concern noted in developmental assessment at 2 years (21.0% vs 16.5% (adjRR 1.08, 95% CI 1.01 to 1.16)). CONCLUSIONS: General anesthesia for cesarean section, irrespective of urgency, is associated with neonatal resuscitation, low Apgar, and neonatal unit admission. Associations were strongest in non-urgent cases and at term. Further evaluation of long-term outcomes is warranted.

Population implications of cessation of IVF during the COVID-19 pandemic.

RESEARCH QUESTION: Discontinuation of IVF cycles has been part of the radical transformation of healthcare provision to enable reallocation of staff and resources to deal with the COVID-19 pandemic. This study sought to estimate the impact of cessation of treatment on individual prognosis and US population live birth rates. DESIGN: Data from 271,438 ovarian stimulation UK IVF cycles was used to model the effect of age as a continuous, yet non-linear, function on cumulative live birth rate. This model was recalibrated to cumulative live birth rates reported for the 135,673 stimulation cycles undertaken in the USA in 2016, with live birth follow-up to October 2018. The effect of a 1-month, 3-month and 6-month shutdown in IVF treatment was calculated as the effect of the equivalent increase in a woman's age, stratified by age group. RESULTS: The average reduction in cumulative live birth rate would be 0.3% (95% confidence interval [CI] 0.3-0.3), 0.8% (95% CI 0.8-0.8) and 1.6% (95% CI 1.6-1.6) for 1-month, 3-month and 6-month shutdowns. This corresponds to a reduction of 369 (95% CI 360-378), 1098 (95% CI 1071-1123) and 2166 (95% CI 2116-2216) live births in the cohort, respectively. Th e greatest contribution to this reduction was from older mothers. CONCLUSIONS: The study demonstrated that the discontinuation of fertility treatment for even 1 month in the USA could result in 369 fewer women having a live birth, due to the increase in patients' age during the shutdown. As a result of reductions in cumulative live birth rate, more cycles may be required to overcome infertility at individual and population levels.

SARS-Cov-2 viral and serological screening of staff in 31 European fertility units.

STUDY QUESTION: What is the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viral presence and seroconversion in staff members in European fertility units prior to recommencement of clinical activity? SUMMARY ANSWER: A large proportion of fertility clinic staff remain susceptible to SARS-CoV-2 with no evidence of seroconversion, indicating that continued comprehensive risk mitigation strategies are essential. WHAT IS KNOWN ALREADY: In response to the coronavirus disease 2019 (COVID-19) pandemic, caused by SARS-CoV-2, routine fertility treatment was temporarily stopped in several European countries. The SARS-CoV-2 prevalence and seroconversion in fertility clinic staff, who are at potentially lower risk than routine healthcare workers, are unknown. STUDY DESIGN SIZE DURATION: This cross-sectional study included 554 staff in 16 European IVF clinics, 13 ultrasound clinics, one diagnostic laboratory and one head office in four European countries (Austria, Denmark, Germany and the UK) between 15 April and 30 June 2020. PARTICIPANTS/MATERIALS SETTING METHODS: There were 554 staff members returning for resumption of clinical activity. Paired nucleic acid amplification tests of oropharyngeal swabs for SARS-CoV-2 and serological testing for SARS-CoV-2 IgG were performed. MAIN RESULTS AND THE ROLE OF CHANCE: Of the 554 staff members tested, 0.19% (95% CI 0.03, 1.10%) had evidence of SARS-CoV-2 as detected by RT-PCR. In contrast, 23 staff members, i.e. 4.15% (95% CI 2.78, 6.15%), had antibodies against SARS-CoV-2, with a wide range of antibody titres. There was no evidence of differences in seroconversion between countries with estimates ranging from 2.78% (95% CI 0.77, 9.58) in Austria to 6.75% (95% CI 4.46, 10.1) for the UK. There was no strong evidence of clustering within the clinics, with 21 of the 30 facilities having no staff members affected (prevalence estimates ranging from 0% to 35%), and one clinic having seven staff members affected (35% (95% CI 18.1%, 56.7%)). The single staff member who tested positive for SARS-CoV-2 virus was in the pre-symptomatic phase and was isolated, with no contacts having evidence of infection on repeat testing. LIMITATIONS REASONS FOR CAUTION: This was a cross-sectional study prior to resumption of clinical activity, with repeat testing not undertaken. WIDER IMPLICATIONS OF THE FINDINGS: The low prevalence of seroconversion of fertility clinic staff highlights the need for continued comprehensive risk mitigation strategies and engagement with national endeavours to identify and isolate new cases and their contacts as we embark on the resumption of fertility services. STUDY FUNDING/COMPETING INTERESTS: The Fertility Partnership funded the study. S.M.N. reports personal fees from Access Fertility, personal fees from Merck, personal fees from Ferring, grants and personal fees from Roche Diagnostics, personal fees from The Fertility Partnership and personal fees from Modern Fertility, outside the submitted work. T.C. reports personal fees from Merck and personal fees from Ferring, outside the submitted work. G.T. reports personal fees from Merck, personal fees from Ferring and personal fees from Roche Diagnostics, outside the submitted work. S.E. and P.S.G. report no conflicts of interest. TRIAL REGISTRATION NUMBER: N/A.

Environmental control programs the emergence of distinct functional ensembles from unconstrained chemical reactions.

Many approaches to the origin of life focus on how the molecules found in biology might be made in the absence of biological processes, from the simplest plausible starting materials. Another approach could be to view the emergence of the chemistry of biology as process whereby the environment effectively directs "primordial soups" toward structure, function, and genetic systems over time. This does not require the molecules found in biology today to be made initially, and leads to the hypothesis that environment can direct chemical soups toward order, and eventually living systems. Herein, we show how unconstrained condensation reactions can be steered by changes in the reaction environment, such as order of reactant addition, and addition of salts or minerals. Using omics techniques to survey the resulting chemical ensembles we demonstrate there are distinct, significant, and reproducible differences between the product mixtures. Furthermore, we observe that these differences in composition have consequences, manifested in clearly different structural and functional properties. We demonstrate that simple variations in environmental parameters lead to differentiation of distinct chemical ensembles from both amino acid mixtures and a primordial soup model. We show that the synthetic complexity emerging from such unconstrained reactions is not as intractable as often suggested, when viewed through a chemically agnostic lens. An open approach to complexity can generate compositional, structural, and functional diversity from fixed sets of simple starting materials, suggesting that differentiation of chemical ensembles can occur in the wider environment without the need for biological machinery.

Designing Algorithms To Aid Discovery by Chemical Robots.

Recently, automated robotic systems have become very efficient, thanks to improved coupling between sensor systems and algorithms, of which the latter have been gaining significance thanks to the increase in computing power over the past few decades. However, intelligent automated chemistry platforms for discovery orientated tasks need to be able to cope with the unknown, which is a profoundly hard problem. In this Outlook, we describe how recent advances in the design and application of algorithms, coupled with the increased amount of chemical data available, and automation and control systems may allow more productive chemical research and the development of chemical robots able to target discovery. This is shown through examples of workflow and data processing with automation and control, and through the use of both well-used and cutting-edge algorithms illustrated using recent studies in chemistry. Finally, several algorithms are presented in relation to chemical robots and chemical intelligence for knowledge discovery.

Miller-Urey Spark-Discharge Experiments in the Deuterium World.

We designed and conducted a series of primordial-soup Miller-Urey style experiments with deuterated gases and reagents to compare the spark-discharge products of a "deuterated world" with the standard reaction in the "hydrogenated world". While the deuteration of the system has little effect on the distribution of amino acid products, significant differences are seen in other regions of the product-space. Not only do we observe about 120 new species, we also see significant differences in their distribution if the two hydrogen isotope worlds are compared. Several isotopologue matches can be identified in both, but a large proportion of products have no equivalent in the corresponding isotope world with ca. 43 new species in the D world and ca. 39 new species in the H world. This shows that isotopic exchange (the addition of only one neutron) may lead to significant additional complexity in chemical space under otherwise identical reaction conditions.

Ethnic differences in male reproductive hormones and relationships with adiposity and insulin resistance in older men.

OBJECTIVES: To assess ethnic differences in male reproductive hormone levels and to determine whether any differences are explained by adiposity, insulin resistance (IR) or comorbidities in older men. DESIGN: Multi-ethnic cross-sectional observational study. PARTICIPANTS: Community dwelling middle-aged and elderly men residing in the UK aged 40-84 years of South Asian (SA; n = 180), White European (WE; n = 328) or African Caribbean (AC; n = 166) origin. OBSERVATIONS: Measured testosterone (T), calculated free T (cFT), sex hormone-binding globulin and LH in SA, WE and AC men along with an assessment of body composition, IR, lifestyle factors and medical conditions. RESULTS: Age-adjusted mean T and cFT levels were lower in SA men when compared to WE and AC men (mean (SEM) T: SA: 14·0 ± 0·4; WE: 17·1 ± 0·3; AC: 17·2 ± 0·5 nmol/l, P < 0·001; cFT: SA: 283 ± 7; WE: 313 ± 5; AC: 314 ± 8 pmol/l, P < 0·006). Compared to WE and AC men, SA men had higher levels of body fat, IR, comorbidities and diabetes. After adjusting for body fat, IR and other confounders, T levels in SA men remained lower than in WE men (P = 0·04) but ethnic differences in cFT became nonsignificant. LH levels were higher in SA than WE men in age-adjusted and fully adjusted models. CONCLUSIONS: T and cFT are lower in SA men than in WE and AC men. Whether ethnic-specific reference ranges for T and cFT might be appropriate in clinical practice requires further investigation. Ethnic differences in cFT, but not T, appear to be, more readily, explained by ethnic differences in adiposity, thus providing insights into potential pathophysiological mechanisms.

Multivariate data analysis and metabolic profiling of artemisinin and related compounds in high yielding varieties of Artemisia annua field-grown in Madagascar.

An improved liquid chromatography-tandem mass spectrometry (LC-MS/MS) protocol for rapid analysis of co-metabolites of A. annua in raw extracts was developed and extensively characterized. The new method was used to analyse metabolic profiles of 13 varieties of A. annua from an in-field growth programme in Madagascar. Several multivariate data analysis techniques consistently show the association of artemisinin with dihydroartemisinic acid. These data support the hypothesis of dihydroartemisinic acid being the late stage precursor to artemisinin in its biosynthetic pathway.

A tutorial review: Metabolomics and partial least squares-discriminant analysis--a marriage of convenience or a shotgun wedding.

The predominance of partial least squares-discriminant analysis (PLS-DA) used to analyze metabolomics datasets (indeed, it is the most well-known tool to perform classification and regression in metabolomics), can be said to have led to the point that not all researchers are fully aware of alternative multivariate classification algorithms. This may in part be due to the widespread availability of PLS-DA in most of the well-known statistical software packages, where its implementation is very easy if the default settings are used. In addition, one of the perceived advantages of PLS-DA is that it has the ability to analyze highly collinear and noisy data. Furthermore, the calibration model is known to provide a variety of useful statistics, such as prediction accuracy as well as scores and loadings plots. However, this method may provide misleading results, largely due to a lack of suitable statistical validation, when used by non-experts who are not aware of its potential limitations when used in conjunction with metabolomics. This tutorial review aims to provide an introductory overview to several straightforward statistical methods such as principal component-discriminant function analysis (PC-DFA), support vector machines (SVM) and random forests (RF), which could very easily be used either to augment PLS or as alternative supervised learning methods to PLS-DA. These methods can be said to be particularly appropriate for the analysis of large, highly-complex data sets which are common output(s) in metabolomics studies where the numbers of variables often far exceed the number of samples. In addition, these alternative techniques may be useful tools for generating parsimonious models through feature selection and data reduction, as well as providing more propitious results. We sincerely hope that the general reader is left with little doubt that there are several promising and readily available alternatives to PLS-DA, to analyze large and highly complex data sets.

A comparison of different chemometrics approaches for the robust classification of electronic nose data.

Accurate detection of certain chemical vapours is important, as these may be diagnostic for the presence of weapons, drugs of misuse or disease. In order to achieve this, chemical sensors could be deployed remotely. However, the readout from such sensors is a multivariate pattern, and this needs to be interpreted robustly using powerful supervised learning methods. Therefore, in this study, we compared the classification accuracy of four pattern recognition algorithms which include linear discriminant analysis (LDA), partial least squares-discriminant analysis (PLS-DA), random forests (RF) and support vector machines (SVM) which employed four different kernels. For this purpose, we have used electronic nose (e-nose) sensor data (Wedge et al., Sensors Actuators B Chem 143:365-372, 2009). In order to allow direct comparison between our four different algorithms, we employed two model validation procedures based on either 10-fold cross-validation or bootstrapping. The results show that LDA (91.56% accuracy) and SVM with a polynomial kernel (91.66% accuracy) were very effective at analysing these e-nose data. These two models gave superior prediction accuracy, sensitivity and specificity in comparison to the other techniques employed. With respect to the e-nose sensor data studied here, our findings recommend that SVM with a polynomial kernel should be favoured as a classification method over the other statistical models that we assessed. SVM with non-linear kernels have the advantage that they can be used for classifying non-linear as well as linear mapping from analytical data space to multi-group classifications and would thus be a suitable algorithm for the analysis of most e-nose sensor data.

Influence of missing values substitutes on multivariate analysis of metabolomics data.

Missing values are known to be problematic for the analysis of gas chromatography-mass spectrometry (GC-MS) metabolomics data. Typically these values cover about 10%-20% of all data and can originate from various backgrounds, including analytical, computational, as well as biological. Currently, the most well known substitute for missing values is a mean imputation. In fact, some researchers consider this aspect of data analysis in their metabolomics pipeline as so routine that they do not even mention using this replacement approach. However, this may have a significant influence on the data analysis output(s) and might be highly sensitive to the distribution of samples between different classes. Therefore, in this study we have analysed different substitutes of missing values namely: zero, mean, median, k-nearest neighbours (kNN) and random forest (RF) imputation, in terms of their influence on unsupervised and supervised learning and, thus, their impact on the final output(s) in terms of biological interpretation. These comparisons have been demonstrated both visually and computationally (classification rate) to support our findings. The results show that the selection of the replacement methods to impute missing values may have a considerable effect on the classification accuracy, if performed incorrectly this may negatively influence the biomarkers selected for an early disease diagnosis or identification of cancer related metabolites. In the case of GC-MS metabolomics data studied here our findings recommend that RF should be favored as an imputation of missing value over the other tested methods. This approach displayed excellent results in terms of classification rate for both supervised methods namely: principal components-linear discriminant analysis (PC-LDA) (98.02%) and partial least squares-discriminant analysis (PLS-DA) (97.96%) outperforming other imputation methods.

A comparative investigation of modern feature selection and classification approaches for the analysis of mass spectrometry data.

Many analytical approaches such as mass spectrometry generate large amounts of data (input variables) per sample analysed, and not all of these variables are important or related to the target output of interest. The selection of a smaller number of variables prior to sample classification is a widespread task in many research studies, where attempts are made to seek the lowest possible set of variables that are still able to achieve a high level of prediction accuracy; in other words, there is a need to generate the most parsimonious solution when the number of input variables is huge but the number of samples/objects are smaller. Here, we compare several different variable selection approaches in order to ascertain which of these are ideally suited to achieve this goal. All variable selection approaches were applied to the analysis of a common set of metabolomics data generated by Curie-point pyrolysis mass spectrometry (Py-MS), where the goal of the study was to classify the Gram-positive bacteria Bacillus. These approaches include stepwise forward variable selection, used for linear discriminant analysis (LDA); variable importance for projection (VIP) coefficient, employed in partial least squares-discriminant analysis (PLS-DA); support vector machines-recursive feature elimination (SVM-RFE); as well as the mean decrease in accuracy and mean decrease in Gini, provided by random forests (RF). Finally, a double cross-validation procedure was applied to minimize the consequence of overfitting. The results revealed that RF with its variable selection techniques and SVM combined with SVM-RFE as a variable selection method, displayed the best results in comparison to other approaches.