Network propagation for GWAS analysis: a practical guide to leveraging molecular networks for disease gene discovery.

MOTIVATION: Genome-wide association studies (GWAS) have enabled large-scale analysis of the role of genetic variants in human disease. Despite impressive methodological advances, subsequent clinical interpretation and application remains challenging when GWAS suffer from a lack of statistical power. In recent years, however, the use of information diffusion algorithms with molecular networks has led to fruitful insights on disease genes. RESULTS: We present an overview of the design choices and pitfalls that prove crucial in the application of network propagation methods to GWAS summary statistics. We highlight general trends from the literature, and present benchmark experiments to expand on these insights selecting as case study three diseases and five molecular networks. We verify that the use of gene-level scores based on GWAS P-values offers advantages over the selection of a set of 'seed' disease genes not weighted by the associated P-values if the GWAS summary statistics are of sufficient quality. Beyond that, the size and the density of the networks prove to be important factors for consideration. Finally, we explore several ensemble methods and show that combining multiple networks may improve the network propagation approach.

Multimodal learning in clinical proteomics: enhancing antimicrobial resistance prediction models with chemical information.

MOTIVATION: Large-scale clinical proteomics datasets of infectious pathogens, combined with antimicrobial resistance outcomes, have recently opened the door for machine learning models which aim to improve clinical treatment by predicting resistance early. However, existing prediction frameworks typically train a separate model for each antimicrobial and species in order to predict a pathogen's resistance outcome, resulting in missed opportunities for chemical knowledge transfer and generalizability. RESULTS: We demonstrate the effectiveness of multimodal learning over proteomic and chemical features by exploring two clinically relevant tasks for our proposed deep learning models: drug recommendation and generalized resistance prediction. By adopting this multi-view representation of the pathogenic samples and leveraging the scale of the available datasets, our models outperformed the previous single-drug and single-species predictive models by statistically significant margins. We extensively validated the multi-drug setting, highlighting the challenges in generalizing beyond the training data distribution, and quantitatively demonstrate how suitable representations of antimicrobial drugs constitute a crucial tool in the development of clinically relevant predictive models. AVAILABILITY AND IMPLEMENTATION: The code used to produce the results presented in this article is available at https://github.com/BorgwardtLab/MultimodalAMR.

Getting personal with epigenetics: towards individual-specific epigenomic imputation with machine learning.

Epigenetic modifications are dynamic mechanisms involved in the regulation of gene expression. Unlike the DNA sequence, epigenetic patterns vary not only between individuals, but also between different cell types within an individual. Environmental factors, somatic mutations and ageing contribute to epigenetic changes that may constitute early hallmarks or causal factors of disease. Epigenetic modifications are reversible and thus promising therapeutic targets for precision medicine. However, mapping efforts to determine an individual's cell-type-specific epigenome are constrained by experimental costs and tissue accessibility. To address these challenges, we developed eDICE, an attention-based deep learning model that is trained to impute missing epigenomic tracks by conditioning on observed tracks. Using a recently published set of epigenomes from four individual donors, we show that transfer learning across individuals allows eDICE to successfully predict individual-specific epigenetic variation even in tissues that are unmapped in a given donor. These results highlight the potential of machine learning-based imputation methods to advance personalized epigenomics.

Machine-Learning-Aided Prediction of Brain Metastases Development in Non-Small-Cell Lung Cancers.

PURPOSE: Non-small-cell lung cancer (NSCLC) shows a high incidence of brain metastases (BM). Early detection is crucial to improve clinical prospects. We trained and validated classifier models to identify patients with a high risk of developing BM, as they could potentially benefit from surveillance brain MRI. METHODS: Consecutive patients with an initial diagnosis of NSCLC from January 2011 to April 2019 and an in-house chest-CT scan (staging) were retrospectively recruited at a German lung cancer center. Brain imaging was performed at initial diagnosis and in case of neurological symptoms (follow-up). Subjects lost to follow-up or still alive without BM at the data cut-off point (12/2020) were excluded. Covariates included clinical and/or 3D-radiomics-features of the primary tumor from staging chest-CT. Four machine learning models for prediction (80/20 training) were compared. Gini Importance and SHAP were used as measures of importance; sensitivity, specificity, area under the precision-recall curve, and Matthew's Correlation Coefficient as evaluation metrics. RESULTS: Three hundred and ninety-five patients compromised the clinical cohort. Predictive models based on clinical features offered the best performance (tuned to maximize recall: sensitivity∼70%, specificity∼60%). Radiomics features failed to provide sufficient information, likely due to the heterogeneity of imaging data. Adenocarcinoma histology, lymph node invasion, and histological tumor grade were positively correlated with the prediction of BM, age, and squamous cell carcinoma histology were negatively correlated. A subgroup discovery analysis identified 2 candidate patient subpopulations appearing to present a higher risk of BM (female patients + adenocarcinoma histology, adenocarcinoma patients + no other distant metastases). CONCLUSION: Analysis of the importance of input features suggests that the models are learning the relevant relationships between clinical features/development of BM. A higher number of samples is to be prioritized to improve performance. Employed prospectively at initial diagnosis, such models can help select high-risk subgroups for surveillance brain MRI.

A historical perspective of biomedical explainable AI research.

The black-box nature of most artificial intelligence (AI) models encourages the development of explainability methods to engender trust into the AI decision-making process. Such methods can be broadly categorized into two main types: post hoc explanations and inherently interpretable algorithms. We aimed at analyzing the possible associations between COVID-19 and the push of explainable AI (XAI) to the forefront of biomedical research. We automatically extracted from the PubMed database biomedical XAI studies related to concepts of causality or explainability and manually labeled 1,603 papers with respect to XAI categories. To compare the trends pre- and post-COVID-19, we fit a change point detection model and evaluated significant changes in publication rates. We show that the advent of COVID-19 in the beginning of 2020 could be the driving factor behind an increased focus concerning XAI, playing a crucial role in accelerating an already evolving trend. Finally, we present a discussion with future societal use and impact of XAI technologies and potential future directions for those who pursue fostering clinical trust with interpretable machine learning models.

Targeted dose enhancement in radiotherapy for breast cancer using gold nanoparticles, part 1: A radiobiological model study.

PURPOSE: Gold nanoparticles (GNPs) are being proposed in combination with radiotherapy to improve tumor control. However, the exact mechanisms underlying GNP radiosensitization are yet to be understood, thus, we present a new approach to estimate the nanoparticle-driven increase in radiosensitivity. METHODS: A stochastic radiobiological model, derived from the Local Effect Model (LEM), was coupled with Monte Carlo simulations to estimate the increase in radiosensitivity produced by the interactions between photons and GNPs at nanometric scale. The model was validated using in vitro survival data of MDA-MB-231 breast cancer cells containing different concentrations of 2 nm diameter GNPs receiving different doses using 160 kVp, 6 MV, and 15 MV photons. A closed analytical formulation of the model was also derived and a study of RBE and TCP behavior was conducted. RESULTS: Results support the increased radiosensitivity due to GNP-driven dose inhomogeneities on a nanometric scale. The model is in good agreement with experimental clonogenic survival assays for 160 kVp, 6 MV, and 15 MV photons. The model suggests a RBE and TCP enhancement when lower energies and lower doses per fraction are used in the presence of GNPs. CONCLUSIONS: The evolution of the local effect model was implemented to assess cellular radiosensitization in the presence of GNPs and then validated with in vitro data. The model provides a useful framework to estimate the nanoparticle-driven radiosensitivity in treatment irradiations and could be applied to real clinical treatment predictions (described in a second part of this paper).

Targeted dose enhancement in radiotherapy for breast cancer using gold nanoparticles, part 2: A treatment planning study.

PURPOSE: In recent years, there has been growing interest in the use of gold nanoparticles (GNPs) combined with radiotherapy to improve tumor control. However, the complex interplay between GNP uptake and dose distribution in realistic clinical treatment are still somewhat unknown. METHODS: The effects of different concentrations of 2 nm diameter GNP, ranging from 0 to 5×105 nanoparticles per tumoral cell, were theoretically investigated. A parametrization of the GNP distribution outside the target was carried out using a Gaussian standard deviation σ, from a zero value, relative to a selective concentration of GNPs inside the tumor volume alone, to 50mm, when GNPs are spatially distributed also in the healthy tissues surrounding the tumor. Treatment simulations of five patients with breast cancer were performed with 6 and 15 MV photons assuming a partial breast irradiation. A closed analytical reformulation of the Local Effect Model coupled with the estimation of local dose deposited around a GNP was validated using an in vitro study for MDA-MB-231 tumoral cells. The expected treatment outcome was quantified in terms of tumor control probability (TCP) and normal tissue complication probability (NTCP) as a function of the spatially varying gold uptake. RESULTS: Breast cancer treatment planning simulations show improved treatment outcomes when GNPs are selectively concentrated in the tumor volume (i.e., σ = 0 mm). In particular, the TCP increases up to 18% for 5×105 nanoparticles per cell in the tumor region depending on the treatment schedules, whereas an improvement of the therapeutic index is observed only for concentrations of about 105 GNPs per tumoral cell and limited spatial distribution in the normal tissue. CONCLUSIONS: The model provides a useful framework to estimate the nanoparticle-driven radiosensitivity in breast cancer treatment irradiation, accounting for the complex interplay between dose and GNP uptake distributions.

"Misura" project: a retrospective survey on the use of 5fluorouracil in the treatment of colorectal cancer in 24 Italian clinical centers.

The "Misura" project is a retrospective survey, with the aim to evaluate how 5FU is used in the treatment of colorectal cancer in clinical practice in Italian oncology departments. Twenty-four centers participated. Patients seen in the second half of 1998 with colorectal cancer and treated with 5FU were analyzed. Observed patients were 664, 45.9% of patients presented metastatic disease. Biochemical modulation with folinic acid and bolus 5FU was the most used schedule (59%). The De Gramont (LV 5FU2) regimen, alone or with other cytotoxic drugs, was the second most chosen schedule (14%). The most frequent side effect observed was gastrointestinal toxicity. No hematological toxicity was demonstrated in 68.8% of patients. Cutaneous toxicity occurred in 21.1% of patients. 5FU is widely used independently by the stage of disease. In palliative treatment a variety of schedules were administered by the Italian centers, lacking a standard therapy. There are very few surveys investigating oncology clinical practice. A larger survey on this issue is auspicable.

Prevalence of common mental disorders in Italy: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD).

OBJECTIVE: To present 1-month, 12-month and lifetime prevalence estimates of mood, anxiety and alcohol disorders in Italy; and the socio-demographic correlates and comorbidity patterns of these estimated disorders. METHOD: A representative random sample of non-institutionalised citizens of Italy aged 18 or older (N=4,712) was interviewed between January 2001 and July 2003, with a weighted response rate of 71.3%. DSM-IV disorders were assessed by lay interviewers using Version 3.0 of the Composite International Diagnostic Interview (CIDI). RESULTS: A total of 11% of respondents reported a lifetime history of any mood disorder, 10.3% any anxiety disorder and 1.3% any alcohol disorder. About 5% reported having an anxiety disorder in the past 12 months compared to 3.3% for any mood disorder and 0.2% for any alcohol disorder. Major depression and specific phobia were the most common mental disorders. Women were twice as likely as men to report a mood disorder and four times as likely as men to report an anxiety disorder, while men were twice as likely as women to report an alcohol disorder. High comorbidity of mood and anxiety disorders was observed. Prevalence estimates were generally lower than in parallel surveys carried out in other Western European countries. CONCLUSION: A high proportion of adults in Italy have a history of mood, anxiety or alcohol disorders. The lower than expected prevalence estimate of alcohol use disorder may be due to under-reporting or to low social harm from alcohol consumption.

Depression and altered quality of life in women with epilepsy of childbearing age.

PURPOSE: To calculate the prevalence of depression in a referral population of women of childbearing age, to define the factors associated with depression, and to assess health-related quality of life (HRQOL) in the same population. METHODS: The 642 consecutive women with epilepsy aged 18-55 years were enrolled by 40 neurologists over an 8-month period and asked to give details on selected demographic and clinical features regarding the disease, any associated clinical condition, and any drug treatment. Depression was diagnosed by using the Hamilton depression scale and HRQOL was measured through the SF-36 form. Demographic, clinical, and therapeutic risk factors for depression were searched for within the study population. RESULTS: Depression (any severity) was present at interview in 242 women, giving a prevalence rate of 37.7%[95% confidence interval (CI), 33.9-41.6]. Mild depression was reported by 18.5% of women, moderate depression by 8.6%, major depression by 10.3%, and severe depression by 0.3%. Factors found to be independently associated with depression (any severity) included treatment of associated conditions [relative risk (RR), 1.5; 95% CI, 1.2-1.8), concurrent disability (RR, 1.3; 95% CI, 1.0-1.6), seizures in the preceding 6 months (RR, 1.4; 95% CI, 1.1-1.7), and being unemployed or a housewife (RR, 1.3; 95% CI, 1.0-1.5). Factors associated with moderate to severe depression included treatment for associated conditions (RR, 2.0; 95% CI, 1.4-2.7), seizures in the preceding 6 months (RR, 1.7; 95% CI, 1.2-2.5), and being unemployed or a housewife (RR, 1.6; 95% CI, 1.1-2.2). Compared with normal women of similar age, patients with epilepsy tended to present lower scores for each HRQOL domain (mostly Role Physical, General Health, Social Functioning, and Role Emotional). However, when the analysis was limited to nondepressed women with epilepsy, any difference disappeared. CONCLUSIONS: Women with epilepsy of childbearing age are at high risk of depression. Factors associated with depression include lack of occupation, the presence of an underlying disabling condition (with treatment), and the severity of epilepsy. Compared with the general population, depressed women have greater impairment of HRQOL with epilepsy, which reflects the physical, social, and emotional implications of the disease.