Accelerated Predictive Healthcare Analytics with Pumas, A High Performance Pharmaceutical Modeling and Simulation Platform

Rackauckas, C. V.; Ma, Y.; Noack, A.; Dixit, V.; Mogensen, P.; Byrne, S.; Maddhashiya, S.; Calderon, J. B. S.; Nyberg, J.; Gobburu, J.; Ivaturi, V.

doi:10.1101/2020.11.28.402297

bioRxiv · 10.1101/2020.11.28.402297

Accelerated Predictive Healthcare Analytics with Pumas, A High Performance Pharmaceutical Modeling and Simulation Platform

Rackauckas, C. V.·Ma, Y.·Noack, A.·Dixit, V.·Mogensen, P.·Byrne, S.·Maddhashiya, S.·Calderon, J. B. S.·Nyberg, J.·Gobburu, J.·Ivaturi, V.

View original source ↗Export metadata ↓

Abstract

Pharmacometric modeling establishes causal quantitative relationships between administered dose, tissue exposures, desired and undesired effects and patients risk factors. These models are employed to de-risk drug development and guide precision medicine decisions. However, pharmacometric tools have not been designed to handle todays heterogeneous big data and complex models. We set out to design a platform that facilitates domain-specific modeling and its integration with modern analytics to foster innovation and readiness in healthcare. Pumas demonstrates estimation methodologies with dramatic performance advances. New ODE solver algorithms, such as coeficient-optimized higher order integrators and new automatic stiffness detecting algorithms which are robust to frequent discontinuities, give rise to a median 4x performance improvement across a wide range of stiff and non-stiff systems seen in pharmacometric applications. These methods combine with JIT compiler techniques, such as statically-sized optimizations and discrete sensitivity analysis via forward-mode automatic differentiation, to further enhance the accuracy and performance of the solving and parameter estimation process. We demonstrate that when all of these techniques are combined with a validated clinical trial dosing mechanism and non-compartmental analysis (NCA) suite, real applications like NLME fitting see a median 81x acceleration while retaining the same accuracy. Meanwhile in areas with less prior software optimization, like optimal experimental design, we see orders of magnitude performance enhancements over competitors. Further, Pumas combines these technical advances with several workflows that are automated and designed to boost productivity of the day-to-day user activity. Together we show a fast pharmacometric modeling framework for next-generation precision analytics.

Explore related subjects

pharmacology and toxicology Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Rackauckas, C. V., Ma, Y., Noack, A., Dixit, V., Mogensen, P., Byrne, S., Maddhashiya, S., Calderon, J. B. S., Nyberg, J., Gobburu, J., Ivaturi, V.. 2020-11-30. Accelerated Predictive Healthcare Analytics with Pumas, A High Performance Pharmaceutical Modeling and Simulation Platform. https://doi.org/10.1101/2020.11.28.402297

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

PCB-Exposed Gut Microbiota Recapitulate Neurobehavioral Deficits and Reprogram Hippocampal Transcriptomic Signatures

Polychlorinated biphenyls (PCBs) are persistent environmental pollutants associated with developmental neurotoxicity and autism spectrum disorder (ASD). However, whether PCB-induced alterations in gut microbiota contribute to neurodevelopmental toxicity independently of direct chemical exposure remains unclear. Here, we investigated whether transplantation of PCB-associated gut microbiota produces early neurobehavioral alterations and disrupts brain molecular pathways in offspring. Gut microbiota from conventional dams exposed to the Fox River PCB mixture (FRM; 6.0 mg/kg/day) or vehicle were transplanted into germ-free recipient females without direct PCB administration. At postnatal day 7 (P7), offspring underwent ultrasonic vocalization (USV) testing, whole-brain RNA sequencing (n = 4/sex/group), and untargeted serum metabolomics. Untargeted serum metabolomics were also performed in donor dams. PCB-associated fecal microbiota transplantation (FMT) significantly reduced offspring USVs, reproducing effects observed following direct maternal PCB exposure. Brain transcriptomics in offspring identified substantially more differentially expressed genes in males (9,339) than females (1,082), including 512 shared genes. Concordantly regulated genes revealed downregulation of neurotransmitter signaling and ion channel pathways alongside upregulation of RNA processing pathways. Male-associated responses involved altered protein synthesis and neuronal signaling, whereas female-associated responses involved DNA repair and extracellular matrix remodeling. Despite these molecular differences, behavioral responses showed no significant sex-dependent effect. Maternal metabolomics revealed increased bile acids and decreased indole-3-propionic acid, highlighting potential AhR-NF{kappa}B and bile acid-associated signaling mechanisms. Offspring metabolomics further identified alterations in lipid-associated metabolites and metabolic profiles. Collectively, these findings show that transferring PCB-exposed microbiota reproduces early neurobehavioral alterations and induces shared and sex-biased molecular responses, identifying the maternal gut microbiome as a candidate mediator linking PCB exposure to developmental neurotoxicity.

pharmacology and toxicology↗

bioRxiv2026

Cisplatin-triggered sensory dysfunction in NER-defective C. elegans involves GABAergic and cholinergic signaling

Cisplatin (CisPt) is a widely used anticancer drug that causes neurotoxicity as a dose-limiting adverse effect. DNA damage contributes to CisPt-induced neurotoxicity, yet the molecular mechanisms involved remain unclear. Here, we aim to identify pathophysiological mechanisms contributing to DNA damage-driven development of sensory neuronal dysfunction using the 3R-compliant model organism Caenorhabditis elegans. To this end, we employed the nucleotide excision repair (NER)-deficient ercc-1 mutant, which shows enhanced CisPt-induced sensory dysfunction as monitored by the chemotaxis assay, to elucidate drug-induced alterations in neurotransmission and transcriptome. We found that CisPt reduces gamma-aminobutyric acid (GABA) levels in the nematode and that GABA supplementation rescues ercc-1 mutants from CisPt-evoked sensory neurotoxicity, pointing to a neuroprotective function of GABA in the context of our study. Moreover, CisPt impairs the morphological integrity of GABAergic neurons particularly in ercc-1 mutants, which again is rescued by GABA substitution. Conversely, the cholinergic agonist carbachol impairs sensory function and exacerbates the neurotoxic effects of CisPt in ercc-1 mutants, indicating detrimental effects resulting from cholinergic excitation. Transcriptome analysis revealed that CisPt treatment stimulates a broard transcriptional response, including pathways involved in the regulation of neuronal communication, including acetylcholine-mediated neurotransmission, with pronounced effects occurring in the absence of ERCC-1. Based on these finding we suggest impaired GABAergic signaling and the loss of neuronal integrity GABAergic neurons as well as excessive cholinergic excitation as key mechanisms linking CisPt-induced DNA damage in NER defective nematodes to sensory dysfunction. Hence, targeting the balance between GABAergic inhibition and cholinergic excitation is suggested to counteract sensory neurotoxicity triggered by CisPt.

pharmacology and toxicology↗

bioRxiv2026

Mathematical model of phenotypic plasticity predicts interconversion is a drug target for neuroblastoma.

Neuroblastoma is a paediatric cancer arising from the sympathetic nervous system as neural crest cells, made up of noradrenergic (NOR/ADRN/N), mesenchymal like (MES/S), and intermediate (I-type/I) cell types. These cells can interconvert between types, contributing to treatment resistance and relapse. We developed a mathematical model to investigate how interconversion patterns affect the efficacy of phenotype-specific inhibitors. We used three ordinary differential equations to describe the dynamics of the three populations, including logistic growth, interconversion, and drug-induced death. Of the 64 possible interconversion patterns, five reduced the target population: two for N cells and three for S cells. N targeting patterns were more effective with higher interconversion rates, while mesenchymal targeting patterns were more effective with lower interconversion rates. Stability analysis revealed that natural tumour growth was always unstable, but during treatment, N-suppressing patterns combined with an S inhibitor gave the most consistently stable origin. Testing multi-stage and cyclic treatment schedules revealed that N-suppressing patterns with an S inhibitor, followed by an S-suppressing pattern with an N inhibitor, achieved more reliable remission. Three of the five patterns of interest correspond to reported biological plausibility, and known compounds exist to induce most of the required transitions, supporting the clinical relevance of these strategies for improving neuroblastoma treatment outcomes.

pharmacology and toxicology↗