Biology subjects

Faghri, F.

Publications and source records attributed to Faghri, F..

4 recordsLinked to original sources

bioRxiv2018

Genetic variability and potential effects on clinical trial outcomes: perspectives in Parkinson’s disease

BackgroundImproper randomization in clinical trials can result in the failure of the trial to meet its primary end-point. The last [~]10 years have revealed that common and rare genetic variants are an important disease factor and sometimes account for a substantial portion of disease risk variance. However, the burden of common genetic risk variants is not often considered in the randomization of clinical trials and can therefore lead to additional unwanted variance between trial arms. We simulated clinical trials to estimate false negative and false positive rates and investigated differences in single variants and mean genetic risk scores (GRS) between trial arms to investigate the potential effect of genetic variance on clinical trial outcomes at different sample sizes.\n\nMethodsSingle variant and genetic risk score analyses were conducted in a clinical trial simulation environment using data from 5851 Parkinsons Disease patients as well as two simulated virtual cohorts based on public data. The virtual cohorts included a GBA variant cohort and a two variant interaction cohort. Data was resampled at different sizes (n = 200-5000 for the Parkinsons Disease cohort) and (n = 50-800 and n = 50-2000 for virtual cohorts) for 1000 iterations and randomly assigned to the two arms of a trial. False negative and false positive rates were estimated using simulated clinical trials, and percent difference in genetic risk score and allele frequency was calculated to quantify disparity between arms.\n\nFindingsSignificant genetic differences between the two arms of a trial are found at all sample sizes. Approximately 90% of the iterations had at least one statistically significant difference in individual risk SNPs between each trial arm. Approximately 10% of iterations had a statistically significant difference between trial arms in polygenic risk score mean or variance. For significant iterations at sample size 200, the average percent difference for mean GRS between trial arms was 130.87%, decreasing to 29.87% as sample size reached 5000. In the GBA only simulations we see an average 18.86% difference in GRS scores between trial arms at n = 50, decreasing to 3.09% as sample size reaches 2000. Balancing patients by genotype reduced mean percent difference in GRS between arms to 36.71% for the main cohort and 2.00% for the GBA cohort at n = 200. When adding a drug effect to the simulations, we found that unbalanced genetics with an effect on the chosen measurable clinical outcome can result in high false negative rates among trials, especially at small sample sizes. At a sample size of n = 50 and a targeted drug effect of -0.5 points in UPDRS per year, we discovered 33.9% of trials resulted in false negatives.\n\nInterpretationsOur data support the hypothesis that within genetically unmatched clinical trials, particularly those below 1000 participants, heterogeneity could confound true therapeutic effects as expected. This is particularly important in the changing environment of drug approvals. Clinical trials should undergo pre-trial genetic adjustment or, at the minimum, post-trial adjustment and analysis for failed trials. Clinical trial arms should be balanced on genetic risk variants, as well as cumulative variant distributions represented by GRS, in order to ensure the maximum reduction in trial arm disparities. The reduction in variance after balancing allows smaller sample sizes to be utilized without risking the large disparities between trial arms witnessed in typical randomized trials. As the cost of genotyping will likely be far less than greatly increasing sample size, genetically balancing trial arms can lead to more cost-effective clinical trials as well as better outcomes.

genomics↗

bioRxiv2018

Parkinson’s disease genetics: identifying novel risk loci, providing causal insights and improving estimates of heritable risk

We performed the largest genome-wide association study of PD to date, involving the analysis of 7.8M SNPs in 37.7K cases, 18.6K UK Biobank proxy-cases, and 1.4M controls. We identified 90 independent genome-wide significant signals across 78 loci, including 38 independent risk signals in 37 novel loci. These variants explained 26-36% of the heritable risk of PD. Tests of causality within a Mendelian randomization framework identified putatively causal genes for 70 risk signals. Tissue expression enrichment analysis suggested that signatures of PD loci were heavily brain-enriched, consistent with specific neuronal cell types being implicated from single cell expression data. We found significant genetic correlations with brain volumes, smoking status, and educational attainment. In sum, these data provide the most comprehensive understanding of the genetic architecture of PD to date by revealing many additional PD risk loci, providing a biological context for these risk factors, and demonstrating that a considerable genetic component of this disease remains unidentified.

genetics↗

bioRxiv2018

Analysis and Prediction of Unplanned Intensive Care Unit Readmission using Recurrent Neural Networks with Long Short-Term Memory

BackgroundUnplanned readmission of a hospitalized patient is an extremely undesirable outcome as the patient may have been exposed to additional risks. The rates of unplanned readmission are, therefore, regarded as an important performance indicator for the medical quality of a hospital and healthcare system. Identifying high-risk patients likely to suffer from readmission before release benefits both the patients and the medical providers. The emergence of machine learning to detect hidden patterns in complex, multi-dimensional datasets provides unparalleled opportunities to develop efficient discharge decision-making support system for physicians.\n\nMethods and FindingsWe used supervised machine learning approaches for ICU readmission prediction. We used machine learning methods on comprehensive, longitudinal clinical data from the MIMIC-III to predict the ICU readmission of patients within 30 days of their discharge. We have utilized recent machine learning techniques such as Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM), by this we have been able incorporate the multivariate features of EHRs and capture sudden fluctuations in chart event features (e.g. glucose and heart rate) that are significant in time series with temporal dependencies, which cannot be properly captured by traditional static models, but can be captured by our proposed deep neural network based model. We incorporate multiple types of features including chart events, demographic, and ICD9 embeddings. Our machine learning models identifies ICU readmissions at a higher sensitivity rate (0.742) and an improved Area Under the Curve (0.791) compared with traditional methods. We also illustrate the importance of each portion of the features and different combinations of the models to verify the effectiveness of the proposed model.\n\nConclusionOur manuscript highlights the ability of machine learning models to improve our ICU decision making accuracy, and is a real-world example of precision medicine in hospitals. These data-driven results enable clinicians to make assisted decisions within their patient cohorts. This knowledge could have immediate implications for hospitals by improving the detection of possible readmission. We anticipate that machine learning models will improve patient counseling, hospital administration, allocation of healthcare resources and ultimately individualized clinical care.

bioinformatics↗

bioRxiv2018

Predicting onset, progression, and clinical subtypes of Parkinson disease using machine learning

BackgroundThe clinical manifestations of Parkinson disease are characterized by heterogeneity in age at onset, disease duration, rate of progression, and constellation of motor versus nonmotor features. Due to these variable presentations, counseling of patients about their individual risks and prognosis is limited. There is an unmet need for predictive tests that facilitate early detection and characterization of distinct disease subtypes as well as improved, individualized predictions of the disease course. The emergence of machine learning to detect hidden patterns in complex, multi-dimensional datasets provides unparalleled opportunities to address this critical need.\n\nMethods and FindingsWe used unsupervised and supervised machine learning approaches for subtype identification and prediction. We used machine learning methods on comprehensive, longitudinal clinical data from the Parkinson Disease Progression Marker Initiative (PPMI) (n=328 cases) to identify patient subtypes and to predict disease progression. The resulting models were validated in an independent, clinically well-characterized cohort from the Parkinson Disease Biomarker Program (PDBP) (n=112 cases). Our analysis distinguished three distinct disease subtypes with highly predictable progression rates, corresponding to slow, moderate and fast disease progressors. We achieved highly accurate projections of disease progression four years after initial diagnosis with an average Area Under the Curve of 0.93 (95% CI: 0.96 {+/-} 0.01 for PDvec1, 0.87 {+/-} 0.03 for PDvec2, and 0.96 {+/-} 0.02 for PDvec3). We have demonstrated robust replication of these findings in the independent validation cohort.\n\nConclusionsThese data-driven results enable clinicians to deconstruct the heterogeneity within their patient cohorts. This knowledge could have immediate implications for clinical trials by improving the detection of significant clinical outcomes that might have been masked by cohort heterogeneity. We anticipate that machine learning models will improve patient counseling, clinical trial design, allocation of healthcare resources and ultimately individualized clinical care.

neuroscience↗