From 0a218346ffed8be90bdde1bce0996ed828d5a31f Mon Sep 17 00:00:00 2001 From: DavAug Date: Sun, 14 Jan 2024 12:44:02 +0000 Subject: [PATCH] Add references and tweaks --- paper.bib | 25 +++++++++++++++++-------- paper.md | 10 +++++----- 2 files changed, 22 insertions(+), 13 deletions(-) diff --git a/paper.bib b/paper.bib index 99120886..43392b82 100644 --- a/paper.bib +++ b/paper.bib @@ -6,7 +6,8 @@ @article{hucka:2003 number={4}, pages={524--531}, year={2003}, - publisher={Oxford University Press} + publisher={Oxford University Press}, + DOI={10.1093/bioinformatics/btg015} } @article{Augustin:2023, @@ -33,6 +34,7 @@ @article{SCHUCK:2015 Title = {Preclinical pharmacokinetic / pharmacodynamic modeling and simulation in the pharmaceutical industry: an IQ consortium survey examining the current landscape}, Volume = {17}, Year = {2015}, + DOI={10.1208/s12248-014-9716-2} } @article{MORGAN:2018, @@ -99,7 +101,8 @@ @article{keizer2013modeling number={6}, pages={1--9}, year={2013}, - publisher={Wiley Online Library} + publisher={Wiley Online Library}, + DOI={10.1038/psp.2013.24} } @article{hosseini2018gpkpdsim, @@ -109,7 +112,8 @@ @article{hosseini2018gpkpdsim volume={45}, pages={259--275}, year={2018}, - publisher={Springer} + publisher={Springer}, + DOI={10.1007/s10928-017-9562-9} } @article{sorzano2021scipion, @@ -120,7 +124,8 @@ @article{sorzano2021scipion number={7}, pages={1169--1178}, year={2021}, - publisher={Springer} + publisher={Springer}, + DOI={10.1007/s11095-021-03065-1} } @article{rackauckas2020accelerated, @@ -129,7 +134,8 @@ @article{rackauckas2020accelerated journal={BioRxiv}, pages={2020--11}, year={2020}, - publisher={Cold Spring Harbor Laboratory} + publisher={Cold Spring Harbor Laboratory}, + DOI={10.1101/2020.11.28.402297} } @article{cole2014maximum, @@ -140,7 +146,8 @@ @article{cole2014maximum number={2}, pages={252--260}, year={2014}, - publisher={Oxford University Press} + publisher={Oxford University Press}, + DOI={10.1093/aje/kwt245} } @article{gelman2014understanding, @@ -151,7 +158,8 @@ @article{gelman2014understanding number={6}, pages={997--1016}, year={2014}, - publisher={Springer} + publisher={Springer}, + DOI={10.1007/s11222-013-9416-2} } @article{augustin2022treatment, @@ -160,7 +168,8 @@ @article{augustin2022treatment journal={bioRxiv}, pages={2022--03}, year={2022}, - publisher={Cold Spring Harbor Laboratory} + publisher={Cold Spring Harbor Laboratory}, + DOI={10.1101/2022.03.19.483454} } @article{clerx2016myokit, diff --git a/paper.md b/paper.md index 508afbc4..6cabc199 100644 --- a/paper.md +++ b/paper.md @@ -20,9 +20,9 @@ bibliography: paper.bib # Summary -[Chi](https://chi.readthedocs.io) is an easy-to-use, open source Python package for the modelling of pharmacokinetics & pharmacodynamics (PKPD). We provide two flexible interfaces to implement PKPD models: 1. an SBML interface, which implements PKPD models based on SBML file specifications [@hucka:2003]; and 2. a general purpose interface that allows users to implement their own, custom PKPD models using Python code. PKPD models instantiated from SBML files automatically implement the administration of custom dosing regimens and the evaluation of parameter sensitivities [@clerx2016myokit]. We also provide a simple framework to extend PKPD models to nonlinear mixed effects (NLME) models, making the simulation of inter-individual variability of treatment responses possible. +[Chi](https://chi.readthedocs.io) is an open source Python package designed for the modelling of treatment responses, with support for implementation, simulation and inference. Supported treatment response models include pharmacokinetic & pharmacodynamic (PKPD) models, physiology-based pharmacokinetic (PBPK) models, quantitative systems pharmacology (QSP) models, and nonlinear mixed effects (NLME) models. The package provides two flexible interfaces to implement single-individual treatment response models: 1. an SBML interface, which implements models based on SBML file specifications [@hucka:2003]; and 2. a general purpose interface that allows users to implement their own, custom models using Python code. Models implemented using SBML files automatically implement routines to administer custom dosing regimens and to evaluate parameter sensitivities using the simulation engine [Myokit](http://myokit.org/) [@clerx2016myokit]. These single-individual treatment response models can then be extending to NLME models, making the simulation of inter-individual variability of treatment responses possible. -In [Chi](https://chi.readthedocs.io), model parameters can be estimated from data using Bayesian inference. We provide a simple interface to estimate posterior distributions of PKPD model parameters and NLME model parameters. [Chi](https://chi.readthedocs.io) also implements filter inference, a novel inference approach which makes the estimation of NLME model parameters from snapshot time series data possible [@Augustin:2023]. +In [Chi](https://chi.readthedocs.io), model parameters can be estimated from data using Bayesian inference. We provide a simple interface to estimate posterior distributions of model parameters for individuals, e.g. PKPD model parameters, or for populations, i.e. NLME model parameters. [Chi](https://chi.readthedocs.io) also implements filter inference, a novel inference approach which makes the estimation of NLME model parameters from snapshot time series data possible [@Augustin:2023]. For the sampling from posterior distributions, [Chi](https://chi.readthedocs.io) uses Markov chain Monte Carlo (MCMC) algorithms implemented in the Python package [PINTS](https://pints.readthedocs.io/en/stable/) [@Clerx:2019]. @@ -30,10 +30,10 @@ Documentation, tutorials and install instructions are available at https://chi.r # Statement of need -PKPD modelling has become an integral part of pharmaceutical research [@SCHUCK:2015; @MORGAN:2018]. In the early phase of drug development, PKPD models help with target and lead identification, and contribute to a (semi-)mechanistic understanding of the relevant pharmacological processes. The modelling results thereafter provide guidance in the transition to the clinical development phase [@LAVE:2016]. During clinical trials, PKPD models help predict the safety, efficacy and treatment response variability of different dosing strategies. More recently, PKPD models are also considered to help with the individualisation of dosing regimens of otherwise difficult-to-administer drugs [@Augustin:20232]. +Treatment response modelling has become an integral part of pharmaceutical research [@SCHUCK:2015; @MORGAN:2018]. In the early phase of drug development, treatment response models help with target and lead identification, and contribute to a (semi-)mechanistic understanding of the relevant pharmacological processes. The modelling results thereafter provide guidance in the transition to the clinical development phase [@LAVE:2016]. During clinical trials, treatment response models help to predict the safety, efficacy and treatment response variability of different dosing strategies. More recently, treatment response models are also being used in the context of model-informed precision dosing, where models help to identify individualised dosing regimens for otherwise difficult-to-administer drugs [@Augustin:20232]. -The most widely used programmes for PKPD modelling include NONMEM [@keizer2013modeling], [Monolix](https://lixoft.com/products/monolix/), and Matlab Simbiology [@hosseini2018gpkpdsim]. Other software packages include Scipion PKPD [@sorzano2021scipion], [PoPy](https://product.popypkpd.com/), Pumas [@rackauckas2020accelerated], and a number of [R libraries](https://cran.r-project.org/web/views/Pharmacokinetics.html). These packages provide an extensive toolkit for PKPD modelling, but are challenging to use for research into novel methodologies for PKPD modelling and inference, as most of them are closed source software packages. The notable exceptions are Scipion PKPD and the R libraries, which share their source code on GitHub. +The most widely used software packages and computer programs for treatment response modelling include NONMEM [@keizer2013modeling], [Monolix](https://lixoft.com/products/monolix/), and Matlab Simbiology [@hosseini2018gpkpdsim]. Other software packages include Scipion PKPD [@sorzano2021scipion], [PoPy](https://product.popypkpd.com/), Pumas [@rackauckas2020accelerated], and a number of [R libraries](https://cran.r-project.org/web/views/Pharmacokinetics.html). These packages provide an extensive toolkit for PKPD modelling. However, for computational and methodological research into PKPD modelling and inference, most of these solutions are difficult to use as their source code is not publicly distributed and or subject to an open-source licenses, concealing the algorithmic details and hindering the methdological development. Notable exceptions are Scipion PKPD and the R libraries, which make their source code publicly available on GitHub. -[Chi](https://chi.readthedocs.io/en/latest/index.html) is an easy-to-use, open-source Python package for the modelling of PKPD processes. It is targeted at PKPD modellers on all levels of programming expertise. For modellers with an interest in methodological research, [Chi](https://chi.readthedocs.io/en/latest/index.html)'s modular, open source framework makes it easy to extend or replace individual components of PKPD models, as well as investigate their advantages and limitations. The straightforward evaluation of log-likelihoods and log-posteriors, and their sensitivities also facilitates the estimation of parametric and structural uncertainty [@cole2014maximum; @gelman2014understanding; @augustin2022treatment]. +[Chi](https://chi.readthedocs.io/en/latest/index.html) is an easy-to-use, open-source Python package for the modelling of treatment responses. It is targeted at PKPD modellers on all levels of programming expertise. For modellers with an interest in methodological research, [Chi](https://chi.readthedocs.io/en/latest/index.html)'s modular, open source framework facilitates extending or replacing individual components of PKPD models. It also enables research into advantages and limitations of different algorithmic and methodological choices. The straightforward evaluation of log-likelihoods and log-posteriors, and their sensitivities also facilitates the estimation of parametric and structural uncertainty [@cole2014maximum; @gelman2014understanding; @augustin2022treatment]. # References \ No newline at end of file