Benchmarking of deep neural networks for predicting personal gene expression from DNA sequence highlights shortcomings.

Deep learning methods have recently become the state of the art in a variety of regulatory genomic tasks1-6, including the prediction of gene expression from genomic DNA. As such, these methods promise to serve as important tools in interpreting the full spectrum of genetic variation observed in personal genomes. Previous evaluation strategies have assessed their predictions of gene expression across genomic regions; however, systematic benchmarking is lacking to assess their predictions across individuals, which would directly evaluate their utility as personal DNA interpreters. We used paired whole genome sequencing and gene expression from 839 individuals in the ROSMAP study7 to evaluate the ability of current methods to predict gene expression variation across individuals at varied loci. Our approach identifies a limitation of current methods to correctly predict the direction of variant effects. We show that this limitation stems from insufficiently learned sequence motif grammar and suggest new model training strategies to improve performance.

PAUSE: principled feature attribution for unsupervised gene expression analysis.

As interest in using unsupervised deep learning models to analyze gene expression data has grown, an increasing number of methods have been developed to make these models more interpretable. These methods can be separated into two groups: post hoc analyses of black box models through feature attribution methods and approaches to build inherently interpretable models through biologically-constrained architectures. We argue that these approaches are not mutually exclusive, but can in fact be usefully combined. We propose PAUSE ( https://github.com/suinleelab/PAUSE ), an unsupervised pathway attribution method that identifies major sources of transcriptomic variation when combined with biologically-constrained neural network models.

Benchmarking of deep neural networks for predicting personal gene expression from DNA sequence highlights shortcomings.

Deep learning methods have recently become the state-of-the-art in a variety of regulatory genomic tasks1-6 including the prediction of gene expression from genomic DNA. As such, these methods promise to serve as important tools in interpreting the full spectrum of genetic variation observed in personal genomes. Previous evaluation strategies have assessed their predictions of gene expression across genomic regions, however, systematic benchmarking is lacking to assess their predictions across individuals, which would directly evaluates their utility as personal DNA interpreters. We used paired Whole Genome Sequencing and gene expression from 839 individuals in the ROSMAP study7 to evaluate the ability of current methods to predict gene expression variation across individuals at varied loci. Our approach identifies a limitation of current methods to correctly predict the direction of variant effects. We show that this limitation stems from insufficiently learnt sequence motif grammar, and suggest new model training strategies to improve performance.

Implementation of a Pediatric Population Health Asthma Program in Academic Medical Center-Affiliated Urban and Suburban Practices.

A tiered pediatric Asthma Population Health Management Program (APHMP), based on evidence-based practices, that differentially targets populations for intervention based on rising risk for high utilization and disease complications was implemented at 6 urban and suburban practices affiliated with an academic medical center. In addition to standard pediatric asthma care, APHMP adds regular administration of the asthma control test (ACT), provider education on performance variation, and monitoring through the electronic health record-based asthma registry. As patients' use of acute health care services and complications increases, APHMP integrates multidisciplinary interventions, including an asthma coach who conducts environmental assessments in addition to addressing social needs, into their primary care. A retrospective cohort study method was used to assess population-level effects on asthma event rates and practice- and provider-level variation from 2017 to 2019. Consistent with well-documented health disparities in pediatric asthma, the analysis demonstrated that patients who were male (odds ratio [OR] = 1.21, 95% confidence interval [CI] = 1.02-1.43), 4-8 years old (OR = 4.91, 95% CI = 3.27-7.37), Spanish speaking (OR = 1.67, 95% CI = 1.54-1.81), from low-income neighborhoods (OR = 1.56, 95% CI = 1.53-2.46), and with ACT <20 (OR = 2.88, 95% CI = 1.97-4.21) had higher odds of having asthma events. Six percent of patients studied were found to be at risk for high health care utilization and disease complications. Study limitations include the absence of a control group, the mixed model data collection approach, and the effects of seasonal variation on asthma events. Future directions include analyzing disease management program outcomes of incorporating an asthma coach into a patient's primary care team and addressing provider-level variation in asthma event rates.

A unique model of the community health worker: the MGH Chelsea Community Health Improvement team.

With current trends in legislation around the delivery of patient care, the role of a community health worker (CHW) is gaining growing and much deserved attention. However, a system needs to be built for any CHW program to be successful and sustainable. This article describes a unique approach to community health work at the Massachusetts General Hospital Chelsea HealthCare Center where a well-integrated CHW model provides support for everyone involved in patient care: patients, providers, the community at large, and the internal CHW staff.