pyDarwin machine learning algorithms application and comparison in nonlinear mixed-effect model selection and optimization.

Forward addition/backward elimination (FABE) has been the standard for population pharmacokinetic model selection (PPK) since NONMEM® was introduced. We investigated five machine learning (ML) algorithms (Genetic algorithm [GA], Gaussian process [GP], random forest [RF], gradient boosted random tree [GBRT], and particle swarm optimization [PSO]) as alternatives to FABE. These algorithms were applied to PPK model selection with a focus on comparing the efficiency and robustness of each of them. All machine learning algorithms included the combination of ML algorithms with a local downhill search. The local downhill search consisted of systematically changing one or two "features" at a time (a one-bit or a two-bit local search), alternating with the ML methods. An exhaustive search (all possible combinations of model features, N = 1,572,864 models) was the gold standard for robustness, and the number of models examined leading prior to identification of the final model was the metric for efficiency.All algorithms identified the optimal model when combined with the two-bit local downhill search. GA, RF, GBRT, and GP identified the optimal model with only a one-bit local search. PSO required the two-bit local downhill search. In our analysis, GP was the most efficient algorithm as measured by the number of models examined prior to finding the optimal (495 models), and PSO exhibited the least efficiency, requiring 1710 unique models before finding the best solution. Additionally, GP was also the algorithm that needed the longest elapsed time of 2975.6 min, in comparison with GA, which only required 321.8 min.

CircRSU1 alleviates LPS-induced human pulmonary microvascular endothelial cell injury by targeting miR-1224-5p/ITGA5 axis.

To investigate the potential functions and regulatory mechanism of circRSU1 on septic acute lung injury (sepsis-ALI) progression. We used lipopolysaccharide (LPS)-stimulated human pulmonary microvascular endothelial cells (HPMECs) to establish the cell model of sepsis-ALI in vitro. qRT-PCR and Western blotting were used for the detection of genes and proteins. The migration and tubulogenesis of HPMECs were assessed by transwell, wound healing, and tube formation assays. Inflammatory factors were detected by ELISA analysis. Cell permeability (PA) was determined by transendothelial resistance (TEER) and fluorescein isothiocyanate (FITC) with transwell assay. The interaction between miR-1224-5p and circRSU1 or ITGA5 (Integrin Subunit Alpha 5) was studied by dual-luciferase reporter and RNA pull-down assays. CircRSU1 expression was decreased after LPS treatment in HPMECs. Functionally, re-expression of circRSU1 in HPMECs could alleviate LPS-induced inflammatory response, the inhibition of cell migration and tube formation and enhancement of cell permeability. Mechanistically, circRSU1 acted as a sponge for miR-1224-5p. LPS treatment enhanced miR-1224-5p expression, and inhibition of miR-1224-5p reversed LPS-evoked HPMEC dysfunction mentioned above. Moreover, miR-1224-5p could abolish the protective effects of circRSU1 on HPMECs. In addition, miR-1224-5p directly targeted ITGA5, and circRSU1 was able to regulate ITGA5 expression via interacting with miR-1224-5p. CircRSU1 could alleviate LPS-induced HPMEC injury by miR-1224-5p/ITGA5 axis, indicating the potential molecular contribution of circRSU1 in sepsis-ALI.

pyDarwin: A Machine Learning Enhanced Automated Nonlinear Mixed-Effect Model Selection Toolbox.

pyDarwin is an open-source Python package for nonlinear mixed-effect model selection. pyDarwin combines machine-learning algorithms and NONMEM to perform a global search for the optimal model in a user-defined model search space. Compared with traditional stepwise search, pyDarwin provides an efficient platform for conducting an objective, robust, less labor-intensive model selection process without compromising model interpretability. In this tutorial, we will begin by introducing the essential components and concepts within the package. Subsequently, we will provide an overview of the pyDarwin modeling workflow and the necessary files needed for model selection. To illustrate the entire process, we will conclude with an example utilizing quetiapine clinical data.

Landscape Analysis of Generic Availability for Oncologic Drugs.

BACKGROUND: Improving generic drug development in oncology is a key long-term goal in providing safe, effective, and affordable care to patients with a diagnosis of cancer in the United States. There are multiple drug and non-drug related variables that may influence generic drug development. To illustrate pertinent associations relevant to generic drug competition in oncology, our study assessed variables that have potentially led to difference in generic competition as compared to drug products in other therapeutic areas, i.e., cardiovascular disease in this case. METHODS: Using a combination of FDA and publicly available data, we categorized individual drug approvals from 1950 to 2021 with either an oncology or cardiovascular indication. Descriptive statistics highlighted the timeline of approval as stratified by indications. Machine learning methodology was used to assess variables associated with abbreviated new drug application (ANDA) availabilities (i.e., generic drug availabilities). Kaplan-Meier analysis with log-rank test compared the difference in the time to approval of first ANDA among products that were off-patent at the time of analysis. A multivariable Cox proportional hazards model with forward selection was used to identify variables (e.g., regulatory recommendation issued, dosage form) that were associated with ANDA availability among products that were off-patent. RESULTS: 434 separate reference listed drugs (RLDs) with varying strengths were identified, 212 (49%) for oncology and 222 (51%) for cardiovascular indications. Compared with cardiovascular products, a greater proportion of RLDs with an oncology indication were approved after 2000 (61% vs. 34%). Also, a smaller proportion of oncologic products had generics (49% vs. 80%). Machine learning methodology revealed RLD age, patent status, product complexity, sales/prescriptions, and regulatory recommendations as variables that were associated with generic availability. Among products off-patent at the time of analysis, the median time from RLD approval to the first ANDA approval was longer for oncologic products compared to cardiovascular products (15.4 years (95% CI 13.8, 17.9) versus 12.3 years (95% CI 10.7, 13.5), p = 0.008). Cox regression analyses identified the variables of product dosage form and regulatory recommendation of requiring patient enrollment for bioequivalence (BE) establishment as being associated with reduced likelihood of ANDA approval for oncologic drugs. CONCLUSION: Oncology indications were found to have a longer time from RLD approval to first ANDA approval compared with cardiovascular drugs. Our work has identified variables that may influence time to ANDA availability, with the requirement of patient enrollment for BE assessment as one important opportunity for future stakeholder engagement and regulatory considerations.

Recruiting Perovskites to Degrade Toxic Trinitrotoluene.

Everybody knows TNT, the most widely used explosive material and a universal measure of the destructiveness of explosions. A long history of use and extensive manufacture of toxic TNT leads to the accumulation of these materials in soil and groundwater, which is a significant concern for environmental safety and sustainability. Reliable and cost-efficient technologies for removing or detoxifying TNT from the environment are lacking. Despite the extreme urgency, this remains an outstanding challenge that often goes unnoticed. We report here that highly controlled energy release from explosive molecules can be accomplished rather easily by preparing TNT-perovskite mixtures with a tailored perovskite surface morphology at ambient conditions. These results offer new insight into understanding the sensitivity of high explosives to detonation initiation and enable many novel applications, such as new concepts in harvesting and converting chemical energy, the design of new, improved energetics with tunable characteristics, the development of powerful fuels and miniaturized detonators, and new ways for eliminating toxins from land and water.

Substantial bulk photovoltaic effect enhancement via nanolayering.

Spontaneous polarization and inversion symmetry breaking in ferroelectric materials lead to their use as photovoltaic devices. However, further advancement of their applications are hindered by the paucity of ways of reducing bandgaps and enhancing photocurrent. By unravelling the correlation between ferroelectric materials' responses to solar irradiation and their local structure and electric polarization landscapes, here we show from first principles that substantial bulk photovoltaic effect enhancement can be achieved by nanolayering PbTiO3 with nickel ions and oxygen vacancies ((PbNiO2)x(PbTiO3)(1-x)). The enhancement of the total photocurrent for different spacings between the Ni-containing layers can be as high as 43 times due to a smaller bandgap and photocurrent direction alignment for all absorption energies. This is due to the electrostatic effect that arises from nanolayering. This opens up the possibility for control of the bulk photovoltaic effect in ferroelectric materials by nanoscale engineering of their structure and composition.

Surface Chemically Switchable Ultraviolet Luminescence from Interfacial Two-Dimensional Electron Gas.

We report intense, narrow line-width, surface chemisorption-activated and reversible ultraviolet (UV) photoluminescence from radiative recombination of the two-dimensional electron gas (2DEG) with photoexcited holes at LaAlO3/SrTiO3. The switchable luminescence arises from an electron transfer-driven modification of the electronic structure via H-chemisorption onto the AlO2-terminated surface of LaAlO3, at least 2 nm away from the interface. The control of the onset of emission and its intensity are functionalities that go beyond the luminescence of compound semiconductor quantum wells. Connections between reversible chemisorption, fast electron transfer, and quantum-well luminescence suggest a new model for surface chemically reconfigurable solid-state UV optoelectronics and molecular sensing.

Ferroelectric Domain Wall Induced Band Gap Reduction and Charge Separation in Organometal Halide Perovskites.

Organometal halide perovskites have been intensely studied in the past 5 years, inspired by their certified high photovoltaic power conversion efficiency. Some of these materials are room-temperature ferroelectrics. The presence of switchable ferroelectric domains in methylammonium lead triiodide, CH3NH3PbI3, has recently been observed via piezoresponse force microscopy. Here, we focus on the structural and electronic properties of ferroelectric domain walls in CH3NH3PbX3 (X = Cl, Br, I). We find that organometal halide perovskites can form both charged and uncharged domain walls due to the flexible orientational order of the organic molecules. The electronic band gaps for domain structures possessing 180 and 90° walls are estimated with density functional theory. It is found that the presence of charged domain walls will significantly reduce the band gap by 20-40%, while the presence of uncharged domain walls has no substantial impact on the band gap. We demonstrate that charged domain walls can serve as segregated channels for the motions of charge carriers. These results highlight the importance of ferroelectric domain walls in hybrid perovskites for photovoltaic applications and suggest a possible avenue for device optimization through domain patterning.

First-Principles Calculation of the Bulk Photovoltaic Effect in CH3NH3PbI3 and CH3NH3PbI(3-x)Cl(x).

Hybrid halide perovskites exhibit nearly 20% power conversion efficiency, but the origin of their high efficiency is still unknown. Here, we compute the shift current, a dominant mechanism of the bulk photovoltaic (PV) effect for ferroelectric photovoltaics, in CH3NH3PbI3 and CH3NH3PbI(3-x)Cl(x) from first-principles. We find that these materials give approximately three times larger shift current PV response to near-IR and visible light than the prototypical ferroelectric photovoltaic BiFeO3. The molecular orientations of CH3NH3⁺ can strongly affect the corresponding PbI3 inorganic frame so as to alter the magnitude of the shift current response. Specifically, configurations with dipole moments aligned in parallel distort the inorganic PbI3 frame more significantly than configurations with near-net-zero dipole, yielding a larger shift current response. Furthermore, we explore the effect of Cl substitution on shift current and find that Cl substitution at the equatorial site induces a larger response than does substitution at the apical site.

Polarization Dependence of Water Adsorption to CH3NH3PbI3 (001) Surfaces.

The instability of organometal halide perovskites when in contact with water is a serious challenge to their feasibility as solar cell materials. Although studies of moisture exposure have been conducted, an atomistic understanding of the degradation mechanism is required. Toward this goal, we study the interaction of water with the (001) surfaces of CH3NH3PbI3 under low and high water concentrations using density functional theory. We find that water adsorption is heavily influenced by the orientation of the methylammonium cations close to the surface. We demonstrate that, depending on methylammonium orientation, the water molecule can infiltrate into the hollow site of the surface and get trapped. Controlling dipole orientation via poling or interfacial engineering could thus enhance its moisture stability. No direct reaction between the water and methylammonium molecules is seen. Furthermore, calculations with an implicit solvation model indicate that a higher water concentration may facilitate degradation through increased lattice distortion.

Structure of Co-Doped Alq3 thin films investigated by grazing incidence X-ray absorption fine structure and Fourier transform infrared spectroscopy.

The structural properties of Co-doped tris(8-hydroxyquinoline)aluminum (Alq(3)) have been studied by grazing incidence X-ray absorption fine structure (GIXAFS) and Fourier transform infrared spectroscopy (FTIR). GIXAFS analysis suggests that there are multivalent Co-Alq(3) complexes and the doped Co atoms tend to locate at the attraction center with respect to N and O atoms and bond with them. The FTIR spectra indicate that the Co atoms interact with the meridional (mer) isomer of Alq(3) rather than forming inorganic compounds.