Forecasting pharmacokinetics (PK) for individual patients is a fundamental problem in clinical pharmacology. One key challenge is that PK models constructed using data from one dosing regimen must predict PK data for different dosing regimen(s). We propose a deep learning approach based on neural ordinary differential equations (neural-ODE) and tested its generalizability against a variety of alternative models. Specifically, we used the PK data from two different treatment regimens of trastuzumab emtansine. The models performed similarly when the training and the test sets come from the same dosing regimen. However, for predicting a new treatment regimen, the neural-ODE model showed substantially better performance. To date, neural-ODE is the most accurate PK model in predicting untested treatment regimens. This study represents the first time neural-ODE has been applied to PK modeling and the results suggest it is a widely applicable algorithm with the potential to impact future studies.
Large-scale population screening and in-home monitoring for patients with Parkinson’s disease (PD) has so far been mainly carried out by traditional healthcare methods and systems. Development of mobile health may provide an independent, future method to detect PD. Current PD detection algorithms will benefit from better generalizability with data collected in real-world situations. In this paper, we report the top-performing smartphone-based method in the recent DREAM Parkinson’s Disease Digital Biomarker Challenge for digital diagnosis of PD. Utilizing real-world accelerometer records, this approach differentiated PD from control subjects with an area under the receiver-operating characteristic curve of 0.87 by 3D augmentation of accelerometer records, a significant improvement over other state-of-the-art methods. This study paves the way for future at-home screening of PD and other neurodegenerative conditions affecting movement.
1 March 2022 in Journal of Medical Internet Research
Background: Monitoring eating is central to the care of many conditions such as diabetes, eating disorders, heart diseases, and dementia. However, automatic tracking of eating in a free-living environment remains a challenge because of the lack of a mature system and large-scale, reliable training set. Objective: This study aims to fill in this gap by an integrative engineering and machine learning effort and conducting a large-scale study in terms of monitoring hours on wearable-based eating detection. Methods: This prospective, longitudinal, passively collected study, covering 3828 hours of records, was made possible by programming a digital system that streams diary, accelerometer, and gyroscope data from Apple Watches to iPhones and then transfers the data to the cloud. Results: On the basis of this data collection, we developed deep learning models leveraging spatial and time augmentation and inferring eating at an area under the curve (AUC) of 0.825 within 5 minutes in the general population. In addition, the longitudinal follow-up of the study design encouraged us to develop personalized models that detect eating behavior at an AUC of 0.872. When aggregated to individual meals, the AUC is 0.951. We then prospectively collected an independent validation cohort in a different season of the year and validated the robustness of the models (0.941 for meal-level aggregation). Conclusions: The accuracy of this model and the data streaming platform promises immediate deployment for monitoring eating in applications such as diabetic integrative care.
Small molecules that inhibit LSD1 (lysine-specific demethylase 1, KDM1A) have been shown to induce abundant fetal hemoglobin (HbF) levels in red blood cells both in vitro and in vivo, therefore potentially serving as potent and cost-effective therapeutics to treat the β-globinopathies, sickle cell disease (SCD) and β-thalassaemia major (TM). However, most LSD1 inhibitors (LSD1i) that induce HbF in vivo are covalent and irreversible, which leads to adverse effects. In this study, we utilized structure-aided drug design to develop potent new reversible LSD1i's leading to robust γ-globin expression in vitro. Moreover, in a mouse model of SCD, oral administration of these novel inhibitors lead to significant HbF elevation and alleviation of multiple features of disease pathology that are the usual consequences of SCD. In addition, we discovered that combined treatment of an LSD1i with a BRD4 degrader (BD-9136) represses the induction of RUNX1 and PU.1, thereby rescuing the erythroid to myeloid lineage conversion that accompanies LSD1i in hematopoiesis. The data indicate that this new generation of LSD1i can effectively induce HbF levels, reduce SCD pathologies, and are well-tolerated by oral administration in SCD mice. We anticipate that the combination of these or related binary compounds offer exciting new therapeutic possibilities for treating SCD and TM.
Nuclear receptor TR4 (NR2C2) was previously shown to bind to the –117 position of the γ-globin gene promoters in vitro, which overlaps the more recently described BCL11 transcription factor A (BCL11A) binding site. The role of TR4 in human γ-globin gene repression has not been extensively characterized in vivo, whereas any relationship between TR4 and BCL11A regulation through the γ-globin promoters is unclear at present. We show here that TR4 and BCL11A competitively bind in vitro to distinct, overlapping sequences, including positions overlapping –117 of the γ-globin promoter. We found that TR4 represses γ-globin transcription and fetal hemoglobin accumulation in vivo in a BCL11A-independent manner. Finally, examination of the chromatin occupancy of TR4 within the β-globin locus, compared with BCL11A, shows that both bind avidly to the locus control region and other sites, but only BCL11A binds to the γ-globin promoters at statistically significant frequency. These data resolve an important discrepancy in the literature and, thus, clarify possible approaches to the treatment of sickle cell disease and β-thalassaemia.
Identifying disease-associated genetic variants remains a key challenge in genomics, especially in small cohorts or for rare and complex mutation types where genome-wide association studies (GWAS) often fall short. We introduce ATLAS, a population-level framework that leverages attention signals from pretrained genomic language models (gLMs) to detect disease-associated genes and loci directly from raw sequences—without requiring explicit variant calls or supervised training. ATLAS first performs gene-level differential attention analysis to prioritize candidate genes, followed by base-level analysis to localize disease-associated regions at single-haplotype resolution. We validate ATLAS on synthetic and β -thalassemia datasets, demonstrating robust performance across diverse allele frequencies (down to 10%), cohort sizes (below 200 individuals per group), and genomic scales. Compared to GWAS, ATLAS achieves higher recall of known loci and captures haplotype-specific signals missed by traditional methods. Cross-model benchmarking further shows that precise localization depends on both model size and pretraining on diverse human genomes. In summary, ATLAS offers a scalable, sequence-native alternative to traditional statistical genetics.