Method development workflow for quantifying protein biomarkers by hybrid LC-MS/MS

Background: Industry-standard guidance on method development and validation of hybrid LC-MS/MS assays for protein biomarkers, particularly on evaluation of parallelism, is lacking. Methods: Using a protein endogenous to humans and mice as a model analyte, a quantitative hybrid LC-MS/MS workflow was developed using a surrogate matrix approach with a recombinant form of the protein as the calibrant. Results: The developed workflow identified a surrogate matrix, established parallelism between the surrogate and authentic matrices and assessed parallelism between the recombinant and authentic forms of the protein. The final method was qualified using precision and accuracy with recovery assessments. Conclusion: The established workflow can be used in future bioanalytical studies to develop effective hybrid LC-MS/MS methods for endogenous protein biomarkers.

Evaluating the Utility of Temporal Compression in High-Resolution Traveling Wave-Based Cyclic Ion Mobility Separations

Ion mobility spectrometry coupled to mass spectrometry (IMS-MS) is slowly becoming a more integral part in omics-based workflows. With the recent technological advancements in IMS-MS instrumentation, particularly those involving traveling wave-based separations, ultralong pathlengths have become readily available in commercial platforms (e.g., Select Series Cyclic IMS from Waters Corporation and MOBIE from MOBILion). However, a tradeoff exists in such ultralong pathlength separations: increasing peak-to-peak resolution at the cost of lower signal intensities and thus poorer sensitivity of measurements. Herein, we explore the utility of temporal compression, where ions are compressed in the time domain, following high-resolution cyclic ion mobility spectrometry-mass spectrometry-based separations on a commercially available, unmodified platform. We assessed temporal compression in the context of various separations including those of reverse sequence peptide isomers, chiral noncovalent complexes, and isotopologues. From our results, we demonstrated that temporal compression improves IMS peak intensities by up to a factor of 4 while only losing ∼5 to 10% of peak-to-peak resolution. Additionally, the improvement in peak quality and signal-to-noise ratio was evident when comparing IMS-MS separations with and without a temporal compression step performed. Temporal compression can readily be implemented in existing traveling wave-based IMS-MS platforms, and our initial proof-of-concept demonstration shows its promise as a tool for improving peak shapes and peak intensities without sacrificing losses in resolution.

Evolving requirements for materials modelling software and underlying method developments: an inventory and future outlook

This European Materials Modelling Council (EMMC) study provides an outline of the survey intent and ambitions, followed by an analysis of the results and a follow up discussion, focused on the future perspectives of the EMMC. The survey covers materials modelling and characterisation communities in both academia and industry. It provides a profile of the surveyed players in these communities and a scaled measure on their usage of computational methodologies. The survey outcomes include: (i) summary views of the recent as well as perceived future trends of materials modelling and its associated fields, with respect to two focus areas surveyed, Model Development and Software, (ii) the main adoption factors and associated bottlenecks for computational methods and software, (iii) the most targeted materials properties and digital twins approaches, and (iv) the wider communities expectations of how EMMC can help facilitate, fulfil and drive further the European Materials Modelling Roadmap to the benefit of the European Commission's (ECs') research and innovation.

Adiabatic state preparation of correlated wave functions with nonlinear scheduling functions and broken-symmetry wave functions

Adiabatic state preparation (ASP) can generate the correlated wave function by simulating the time evolution of wave function under the time-dependent Hamiltonian that interpolates the Fock operator and the full electronic Hamiltonian. However, ASP is inherently unsuitable for studying strongly correlated systems, and furthermore practical computational conditions for ASP are unknown. In quest for the suitable computational conditions for practical applications of ASP, we performed numerical simulations of ASP in the potential energy curves of N₂, BeH₂, and in the C_2v quasi-reaction pathway of the Be atom insertion to the H₂ molecule, examining the effect of nonlinear scheduling functions and the ASP with broken-symmetry wave functions with the S² operator as the penalty term, contributing to practical applications of quantum computing to quantum chemistry. Eventually, computational guidelines to generate the correlated wave functions having the square overlap with the complete-active space self-consistent field wave function close to unity are discussed.

Language models can learn complex molecular distributions

Deep generative models of molecules have grown immensely in popularity, trained on relevant datasets, these models are used to search through chemical space. The downstream utility of generative models for the inverse design of novel functional compounds, depends on their ability to learn a training distribution of molecules. The most simple example is a language model that takes the form of a recurrent neural network and generates molecules using a string representation. Since their initial use, subsequent work has shown that language models are very capable, in particular, recent research has demonstrated their utility in the low data regime. In this work, we investigate the capacity of simple language models to learn more  complex distributions of molecules. For this purpose, we introduce several challenging generative modeling tasks by compiling larger, more complex distributions of molecules and we evaluate the ability of language models on each task. The results demonstrate that language models are powerful generative models, capable of adeptly learning complex molecular distributions. Language models can accurately generate: distributions of the highest scoring penalized LogP molecules in ZINC15, multi-modal molecular distributions as well as the largest molecules in PubChem. The results highlight the limitations of some of the most popular and recent graph generative models- many of which cannot scale to these molecular distributions.

Development and Validation of an HPLC-UV Method for the Quantification of 4'-Hydroxydiclofenac Using Salicylic Acid: Future Applications for Measurement of In Vitro Drug-Drug Interaction in Rat Liver Microsomes

Salicylic acid is a key compound in nonsteroidal anti-inflammatory drugs that has been recently used for preventing the risk of hospitalization and death among COVID-19 patients and in preventing colorectal cancer (CRC) by suppressing two key proteins. Understanding drug−drug interaction pathways prevent the occurrence of adverse drug reactions in clinical trials. Drug−drug interactions can result in the variation of the pharmacodynamics and pharmacokinetic of the drug. Inhibition of the Cytochrome P450 enzyme activity leads to the withdrawal of the drug from the market. The aim of this paper was to develop and validate an HPLC-UV method for the quantification of 4′-hydroxydiclofenac as a CYP2C9 metabolite using salicylic acid as an inhibitor in rat liver microsomes. A CYP2C9 assay was developed and validated on the reversed phase C18 column (SUPELCO 25 cm × 4.6 mm × 5 µm) using a low-pressure gradient elution programming at T = 30 °C, a wavelength of 282 nm, and a flow rate of 1 mL/min. 4′-hydroxydiclofenac demonstrated a good linearity (R2 > 0.99), good reproducibility, low detection, and quantitation limit, and the inter and intra-day precision met the ICH guidelines (<15%). 4′-hydroxydiclofenac was stable for three days and showed an acceptable accuracy and recovery (80−120%) within the ICH guidelines in a rat liver microsome sample. This method will be beneficial for future applications of the in vitro inhibitory effect of salicylic acid on the CYP2C9 enzyme activity in rat microsomes and the in vivo administration of salicylic acid in clinical trials.

Computational methods to simulate molten salt thermophysical properties

Molten salts are important thermal conductors used in molten salt reactors and solar applications. To use molten salts safely, accurate knowledge of their thermophysical properties is necessary. However, it is experimentally challenging to measure these properties and a comprehensive evaluation of the full chemical space is unfeasible. Computational methods provide an alternative route to access these properties. Here, we summarize the developments in methods over the last 70 years and cluster them into three relevant eras. We review the main advances and limitations of each era and conclude with an optimistic perspective for the next decade, which will likely be dominated by emerging machine learning techniques. This article is aimed to help researchers in peripheral scientific domains understand the current challenges of molten salt simulation and identify opportunities to contribute.

Liquid Chromatography High Resolution Mass Spectrometry in Forensic Toxicology: What Are the Specifics of Method Development, Validation and Quality Assurance for Comprehensive Screening Approaches?

The use of High Resolution Mass Spectrometry (HRMS) has increased over the past decade in clinical and forensic toxicology, especially for comprehensive screening approaches. Despite this, few guidelines of this field have specifically addressed HRMS issues concerning compound identification, validation, measurement uncertainty and quality assurance. To fully implement this technique, certainly in an era in which the quality demands for laboratories are ever increasing due to various norms (e.g. the International Organization for Standardization's ISO 17025), these specific issues need to be addressed. This manuscript reviews 26 HRMS-based methods for qualitative systematic toxicological analysis (STA) published between 2011 and 2021. Key analytical data such as samples matrices, analytical platforms, numbers of analytes and employed mass spectral reference databases/libraries as well as the studied validation parameters are summarized and discussed. The article further includes a critical review of targeted and untargeted data acquisition approaches, available HRMS reference databases and libraries as well as current guidelines for HRMS data interpretation with a particular focus on identification criteria. Moreover, it provides an overview on current recommendations for the validation and determination measurement uncertainty of qualitative methods. Finally, the article aims to put forward suggestions for method development, compound identification, validation experiments to be performed, and adequate determination of measurement uncertainty for this type of wide-range qualitative HRMS-based methods.

Cryo-Electron Microscopy: the field of 1,000+ methods

Cryo-Electron Microscopy (CryoEM) is currently a well-established method to elucidate a biological macromolecule's three-dimensional (3D) structure. Its success is due to technological and methodological advances in several fronts: sample preparation, electron optics and detection, image acquisition, image processing, and map interpretation. The first methods started in the late 1960s and, since then, new methods on all fronts have continuously been published, maturating the field as we know it now. In terms of publications, we can distinguish several periods, witnessing a substantial acceleration of methodological publications in recent years, pointing out to an increased interest in the domain. On the other hand, this accelerated increase of methods development may confuse practitioners about which method they should be using (and how) and highlight the importance of paying attention to establishing best practices for methods reporting and usage. In this paper, we analyze the trends identified in over 1,000 methodological papers. Our focus is primarily on computational image processing methods. However, our list also covers some aspects of sample preparation and image acquisition. Several interesting ideas stem out from this study: 1) Single Particle Analysis (SPA) has largely accelerated in the last decade and sample preparation methods in the last five years; 2) Electron Tomography is not yet in a rapidly growing phase, but it is foreseeable that it will soon be; 3) the work horses of SPA are 3D classification, 3D reconstruction, and 3D alignment, and there have been many papers on these topics, which are not considered to be solved yet, but ever improving; and 4) since the resolution revolution, atomic modelling has also caught on as a hot topic.

Strain Characterisation for Measuring Bioefficacy of ITNs Treated with Two Active Ingredients (Dual-AI ITNs): Developing a Robust Protocol by Building Consensus

Durability monitoring of insecticide-treated nets (ITNs) containing a pyrethroid in combination with a second active ingredient (AI) must be adapted so that the insecticidal bioefficacy of each AI can be monitored independently. An effective way to do this is to measure rapid knock down of a pyrethroid-susceptible strain of mosquitoes to assess the bioefficacy of the pyrethroid component and to use a pyrethroid-resistant strain to measure the bioefficacy of the second ingredient. To allow robust comparison of results across tests within and between test facilities, and over time, protocols for bioefficacy testing must include either characterisation of the resistant strain, standardisation of the mosquitoes used for bioassays, or a combination of the two. Through a series of virtual meetings, key stakeholders and practitioners explored different approaches to achieving these goals. Via an iterative process we decided on the preferred approach and produced a protocol consisting of characterising mosquitoes used for bioefficacy testing before and after a round of bioassays, for example at each time point in a durability monitoring study. We present the final protocol and justify our approach to establishing a standard methodology for durability monitoring of ITNs containing pyrethroid and a second AI.

RAVAQ: An integrative pipeline from quality control to region-based rare variant association analysis

Next-generation sequencing technologies have opened up the possibility to sequence large samples of cases and controls to test for association with rare variants. To limit cost and increase sample sizes, data from controls could be used in multiple studies and might thus be generated on different sequencing platforms. This could pose some problems of comparability between cases and controls due to batch effects that could be confounding factors, leading to false-positive association signals. To limit batch effects and ensure comparability of datasets, stringent quality controls are required. We propose an integrative five-steps pipeline, RAVAQ, that (a) performs a specific three-step quality control taking into account the case-control status to ensure data comparability, (b) selects qualifying variants as defined by the user, and (c) performs rare variant association tests per genomic region. The RAVAQ pipeline is wrapped in an R package. It is user-friendly and flexible in its arguments to adapt to the specificity of each research project. We provide examples showing how RAVAQ improves rare variant association tests. The default RAVAQ quality control outperformed the widely used Variant Quality Score Recalibration method, removing inflation due to spurious signals. RAVAQ is open source and freely available at https://gitlab.com/gmarenne/ravaq.

Proteomic-driven identification of short open reading frame-encoded peptides

Accumulating evidence has shown that a large number of short open reading frames (sORFs) also have the ability to encode proteins. The discovery of sORFs opens up a new research area, leading to the identification and functional study of sORF encoded peptides (SEPs) at the omics level. Besides bioinformatics prediction and ribosomal profiling, mass spectrometry (MS) has become a significant tool as it directly detects the sequence of SEPs. Though MS-based proteomics methods have proved to be effective for qualitative and quantitative analysis of SEPs, the detection of SEPs is still a great challenge due to their low abundance and short sequence. To illustrate the progress in method development, we described and discussed the main steps of large-scale proteomics identification of SEPs, including SEP extraction and enrichment, MS detection, data processing and quality control, quantification, and function prediction and validation methods. This article is protected by copyright. All rights reserved.

Fast and Sensitive Quantification of AccQ-Tag Derivatized Amino Acids and Biogenic Amines by UHPLC-UV Analysis from Complex Biological Samples

Metabolomic analysis of different body fluids bears high importance in medical sciences. Our aim was to develop and validate a fast UHPLC-UV method for the analysis of 33 amino acids and biogenic amines from complex biological samples. AccQ-Tag derivatization was conducted on target molecules and the derivatized targets were analyzed by UHPLC-UV. The detection of the analytes was carried out with UV analysis and by Selected Reaction Monitoring (SRM)-based targeted mass spectrometry. The method was validated according to the FDA guidelines. Serum and non-stimulated tear samples were collected from five healthy individuals and the samples were analyzed by the method. The method was successfully validated with appropriate accuracy and precision for all 33 biomolecules. A total of 29 analytes were detected in serum samples and 26 of them were quantified. In the tears, 30 amino acids and biogenic amines were identified and 20 of them were quantified. The developed and validated UHPLC-UV method enables the fast and precise analysis of amino acids and biogenic amines from complex biological samples.

Computational screening methodology identifies effective solvents for CO2 capture

Carbon capture and storage technologies are projected to increasingly contribute to cleaner energy transitions by significantly reducing CO₂ emissions from fossil fuel-driven power and industrial plants. The industry standard technology for CO₂ capture is chemical absorption with aqueous alkanolamines, which are often being mixed with an activator, piperazine, to increase the overall CO₂ absorption rate. Inefficiency of the process due to the parasitic energy required for thermal regeneration of the solvent drives the search for new tertiary amines with better kinetics. Improving the efficiency of experimental screening using computational tools is challenging due to the complex nature of chemical absorption. We have developed a novel computational approach that combines kinetic experiments, molecular simulations and machine learning for the in silico screening of hundreds of prospective candidates and identify a class of tertiary amines that absorbs CO₂ faster than a typical commercial solvent when mixed with piperazine, which was confirmed experimentally.

Systematic optimization of targeted and multiplexed MS-based screening workflows for protein biomarkers

Background: The capability of targeted MS-based methods to simultaneously measure multiple analytes with high selectivity and sensitivity greatly facilitates the discovery and quantitation of novel biomarkers. However, the complexity of biological samples is a major bottleneck that requires extensive sample preparation. Results: This paper reports a generic workflow to optimize surrogate peptide-based protein biomarker screening for seven human proteins in a multiplexed manner without the need for any specific affinity reagents. Each step of the sample processing and LC-MS methods is systematically assessed and optimized for better analytical performance. Conclusion: The established method is used for the screening of multiple myeloma patient samples to determine which proteins could be robustly measured and serve as potential biomarkers of the disease.

Evaluating the Effects of Modified Windscreens on Organic Vapor Monitor Performance

Passive sampling using diffusive samplers has become popular as a convenient means of occupational compliance sampling for volatile organic compounds (VOCs). However, diffusive samplers possess sensitivity limitations when sampling low concentrations and for short durations. To reduce these limitations, our research team has been developing a novel method of sample recovery called photothermal desorption (PTD), which uses high energy visible light pulses to desorb analytes from sampling media. Newly designed passive samplers that will use PTD will be equipped with windscreens in a similar design with the 3M OVM. In a preliminary design effort, the present work sought to find a suitable, windscreen for future use in a PTD-compatible diffusive sampler prototype that would be similar to those found in commercially available diffusive samplers. To do so, 2 stainless steel windscreens (wire diameters 0.015″ and 0.0055″ respectively) were compared to a standard windscreen by exposing modified (ie, steel mesh installed) and non-modified 3M OVM samplers to 3 analytes. To mimic in-field conditions, each sampler was exposed to analyte concentrations at their short-term and personal exposure limits (STELs and PELs). From these comparisons, it was determined that the 0.0055″ mesh was most similar to the standard windscreen in contributing to sample collection based on the uptake and concentration determinations for each analyte and concentration.

Development and Validation of a Reliable UHPLC-MS/MS Method for Simultaneous Quantification of Macrocyclic Lactones in Bovine Plasma

A fast, accurate and reliable ultra-high performance liquid chromatography–tandem mass spectrometry (UHPLC-MS/MS) method was developed for simultaneous quantification of ivermectin (IVER), doramectin (DORA), and moxidectin (MOXI) in bovine plasma. A priority for sample preparation was the eradication of possible infectious diseases to avoid travel restrictions. The sample preparation was based on protein precipitation using 1% formic acid in acetonitrile, followed by Ostro^® 96-well plate pass-through sample clean-up. The simple and straightforward procedure, along with the short analysis time, makes the current method unique and suitable for a large set of sample analyses per day for PK studies. Chromatographic separation was performed using an Acquity UPLC HSS-T3 column, with 0.01% acetic acid in water and methanol, on an Acquity H-Class ultra-high performance liquid chromatograph (UHPLC) system. The MS/MS instrument was a Xevo TQ-S^® mass spectrometer, operating in the positive electrospray ionization mode and two multiple reaction monitoring (MRM) transitions were monitored per component. The MRM transitions of m/z 897.50 > 753.4 for IVER, m/z 921.70 > 777.40 for DORA and m/z 640.40 > 123.10 for MOXI were used for quantification. The method validation was performed using matrix-matched calibration curves in a concentration range of 1 to 500 ng/mL. Calibration curves fitted a quadratic regression model with 1/x2 weighting (r ≥ 0.998 and GoF ≤ 4.85%). Limits of quantification (LOQ) values of 1 ng/mL were obtained for all the analytes, while the limits of detection (LOD) were 0.02 ng/mL for IVER, 0.03 ng/mL for DORA, and 0.58 ng/mL for MOXI. The results of within-day (RSD < 6.50%) and between-day (RSD < 8.10%) precision and accuracies fell within acceptance ranges. No carry-over and no peak were detected in the UHPLC-MS/MS chromatogram of blank samples showing good specificity of the method. The applicability of the developed method was proved by an analysis of the field PK samples.

Simultaneous Quantification of Chloramphenicol, Thiamphenicol, Florfenicol, and Florfenicol Amine in Animal and Aquaculture Products Using Liquid Chromatography-Tandem Mass Spectrometry

The accumulation of antimicrobial residues in edible animal products and aquaculture products could pose health concerns to unsuspecting consumers. Hence, this study aimed to develop a validated method for simultaneous quantification of chloramphenicol (CAP), thiamphenicol (TAP), florfenicol (FF), and florfenicol amine (FFA) in beef, pork, chicken, shrimp, eel, and flatfish using a quick, easy, cheap, effective, rugged, and safe (QuEChERS) extraction method coupled with liquid chromatography-tandem mass spectrometry (LC-MS/MS). Primary-secondary amine (PSA) and MgSO4 were used for sample purification. The analytes were separated on a reversed-phase analytical column. The coefficients of determination for the linear matrix-matched calibration curves were ≥0.9941. Recovery rates ranged between 64.26 and 116.51% for the four analytes with relative standard deviations (RSDs) ≤ 18.05%. The calculated limits of detection (LODs) and limits of quantification (LOQs) were 0.005-3.1 and 0.02-10.4 μg/kg, respectively. The developed method was successfully applied for monitoring samples obtained from local markets in Seoul, Republic of Korea. The target residues were not detected in any tested matrix. The designed method was versatile, sensitive, and proved suitable for quantifying residues in animal-derived products.

SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects

Machine-learned force fields combine the accuracy of ab initio methods with the efficiency of conventional force fields. However, current machine-learned force fields typically ignore electronic degrees of freedom, such as the total charge or spin state, and assume chemical locality, which is problematic when molecules have inconsistent electronic states, or when nonlocal effects play a significant role. This work introduces SpookyNet, a deep neural network for constructing machine-learned force fields with explicit treatment of electronic degrees of freedom and nonlocality, modeled via self-attention in a transformer architecture. Chemically meaningful inductive biases and analytical corrections built into the network architecture allow it to properly model physical limits. SpookyNet improves upon the current state-of-the-art (or achieves similar performance) on popular quantum chemistry data sets. Notably, it is able to generalize across chemical and conformational space and can leverage the learned chemical insights, e.g. by predicting unknown spin states, thus helping to close a further important remaining gap for today's machine learning models in quantum chemistry.

Variant calling: Considerations, practices, and developments

The success of many clinical, association, or population genetics studies critically relies on properly performed variant calling step. The variety of modern genomics protocols, techniques, and platforms makes our choices of methods and algorithms difficult and there is no "one size fits all" solution for study design and data analysis. In this review, we discuss considerations that need to be taken into account while designing the study and preparing for the experiments. We outline the variety of variant types that can be detected using sequencing approaches and highlight some specific requirements and basic principles of their detection. Finally, we cover interesting developments that enable variant calling for a broad range of applications in the genomics field. We conclude by discussing technological and algorithmic advances that have the potential to change the ways of calling DNA variants in the nearest future.

Artificial intelligence-enhanced quantum chemical method with broad applicability

High-level quantum mechanical (QM) calculations are indispensable for accurate explanation of natural phenomena on the atomistic level. Their staggering computational cost, however, poses great limitations, which luckily can be lifted to a great extent by exploiting advances in artificial intelligence (AI). Here we introduce the general-purpose, highly transferable artificial intelligence-quantum mechanical method 1 (AIQM1). It approaches the accuracy of the gold-standard coupled cluster QM method with high computational speed of the approximate low-level semiempirical QM methods for the neutral, closed-shell species in the ground state. AIQM1 can provide accurate ground-state energies for diverse organic compounds as well as geometries for even challenging systems such as large conjugated compounds (fullerene C60) close to experiment. This opens an opportunity to investigate chemical compounds with previously unattainable speed and accuracy as we demonstrate by determining geometries of polyyne molecules-the task difficult for both experiment and theory. Noteworthy, our method's accuracy is also good for ions and excited-state properties, although the neural network part of AIQM1 was never fitted to these properties.

A shortcut to the thermodynamic limit for quantum many-body calculations of metals

Computationally efficient and accurate quantum mechanical approximations to solve the many-electron Schrödinger equation are crucial for computational materials science. Methods such as coupled cluster theory show potential for widespread adoption if computational cost bottlenecks can be removed. For example, extremely dense k-point grids are required to model long-range electronic correlation effects, particularly for metals. Although these grids can be made more effective by averaging calculations over an offset (or twist angle), the resultant cost in time for coupled cluster theory is prohibitive. We show here that a single special twist angle can be found using the transition structure factor, which provides the same benefit as twist averaging with one or two orders of magnitude reduction in computational time. We demonstrate that this not only works for metal systems but also is applicable to a broader range of materials, including insulators and semiconductors.

Development and validation of ultra performance liquid chromatography-tandem mass spectrometry method for the simultaneous estimation of dolutegravir, lamivudine and tenofovir in bulk and tablet dosage form

A simple, selective and rapid ultra performance liquid chromatography-tandem mass spectrometry method was developed and validated for the simultaneous estimation of dolutegravir, lamivudine and tenofovir in bulk and tablet dosage form. Chromatographic separation was attained on Acquity Ethylene Bridged Hybrid (BEH) C18 column (50 × 2.1 mm, 3.5 µm), using a mixture of acetonitrile and 0.1% formic acid in water (60:40, v/v) as a mobile phase at a flow rate of 0.12 mL/min. The total run time of analysis was 3.5 min. The analytes were detected using tandem mass spectrometry, operating in positive ionization and multiple reaction monitoring modes. The method's linearity was determined to be in the range of 10-150 ng/mL with r² > 0.99. The proposed method was validated as per the International Council for Harmonization (ICH) guidelines, and the results were found well within the acceptance limits. The method was successfully applied for the simultaneous quantification of all the three analytes in the combined tablet dosage form.

Augmenting zero-Kelvin quantum mechanics with machine learning for the prediction of chemical reactions at high temperatures

The prediction of temperature effects from first principles is computationally demanding and typically too approximate for the engineering of high-temperature processes. Here, we introduce a hybrid approach combining zero-Kelvin first-principles calculations with a Gaussian process regression model trained on temperature-dependent reaction free energies. We apply this physics-based machine-learning model to the prediction of metal oxide reduction temperatures in high-temperature smelting processes that are commonly used for the extraction of metals from their ores and from electronics waste and have a significant impact on the global energy economy and greenhouse gas emissions. The hybrid model predicts accurate reduction temperatures of unseen oxides, is computationally efficient, and surpasses in accuracy computationally much more demanding first-principles simulations that explicitly include temperature effects. The approach provides a general paradigm for capturing the temperature dependence of reaction free energies and derived thermodynamic properties when limited experimental reference data is available.