Pruning the Way to Reliable Policies: A Multi-Objective Deep Q-Learning Approach to Critical Care.

Medical treatment decisions inherently involve a series of sequential choices, each informed by the outcomes of preceding decisions. This process closely aligns with the principles of reinforcement learning (RL), which also focuses on sequential decisions aimed at maximizing cumulative rewards. Consequently, RL holds significant promise for developing data-driven treatment plans. However, a major challenge in applying RL within medical contexts lies in the sparse nature of the rewards, which are primarily based on mortality outcomes. This sparsity can reduce the stability of offline estimates, posing a significant hurdle in fully utilizing RL for medical decision-making. In this work, we introduce a deep Q-learning approach able to obtain more reliable critical care policies. This method integrates relevant but noisy intermediate biomarker signals into the reward specification without compromising the optimization of the main outcome of interest (e.g., patient survival). We achieve this by first pruning the action space based on all available rewards, and then training a final model based on the (sparse) main reward, while only choosing actions available within the pruned action space. By disentangling sparse rewards and frequently measured reward proxies through action pruning, potential distortions of the main objective are minimized, all while enabling the extraction of valuable information from intermediate signals that can guide the learning process. We evaluate our method in both off-policy and offline settings using simulated environments and real health records of patients in intensive care units. Our empirical results indicate that our method outperforms common offline RL methods such as conservative Q-learning and batch-constrained deep Q-learning. Our work is a step towards developing reliable policies by effectively harnessing the wealth of available information in data-intensive critical care environments.

N4BP1 coordinates ubiquitin-dependent crosstalk within the IκB kinase family to limit Toll-like receptor signaling and inflammation.

The ubiquitin-binding endoribonuclease N4BP1 potently suppresses cytokine production by Toll-like receptors (TLRs) that signal through the adaptor MyD88 but is inactivated via caspase-8-mediated cleavage downstream of death receptors, TLR3, or TLR4. Here, we examined the mechanism whereby N4BP1 limits inflammatory responses. In macrophages, deletion of N4BP1 prolonged activation of inflammatory gene transcription at late time points after TRIF-independent TLR activation. Optimal suppression of inflammatory cytokines by N4BP1 depended on its ability to bind polyubiquitin chains, as macrophages and mice-bearing inactivating mutations in a ubiquitin-binding motif in N4BP1 displayed increased TLR-induced cytokine production. Deletion of the noncanonical IκB kinases (ncIKKs), Tbk1 and Ikke, or their adaptor Tank phenocopied N4bp1 deficiency and enhanced macrophage responses to TLR1/2, TLR7, or TLR9 stimulation. Mechanistically, N4BP1 acted in concert with the ncIKKs to limit the duration of canonical IκB kinase (IKKα/ß) signaling. Thus, N4BP1 and the ncIKKs serve as an important checkpoint against over-exuberant innate immune responses.

How Real-World Data Can Facilitate the Development of Precision Medicine Treatment in Psychiatry.

Precision medicine has the ambition to improve treatment response and clinical outcomes through patient stratification and holds great potential for the treatment of mental disorders. However, several important factors are needed to transform current practice into a precision psychiatry framework. Most important are 1) the generation of accessible large real-world training and test data including genomic data integrated from multiple sources, 2) the development and validation of advanced analytical tools for stratification and prediction, and 3) the development of clinically useful management platforms for patient monitoring that can be integrated into health care systems in real-life settings. This narrative review summarizes strategies for obtaining the key elements-well-powered samples from large biobanks integrated with electronic health records and health registry data using novel artificial intelligence algorithms-to predict outcomes in severe mental disorders and translate these models into clinical management and treatment approaches. Key elements are massive mental health data and novel artificial intelligence algorithms. For the clinical translation of these strategies, we discuss a precision medicine platform for improved management of mental disorders. We use cases to illustrate how precision medicine interventions could be brought into psychiatry to improve the clinical outcomes of mental disorders.

A Computational Study of the Electronic Energy and Charge Transfer Rates and Pathways in the Tetraphenyldibenzoperiflanthene/Fullerene Interfacial Dyad.

The electronic transition rates and pathways underlying interfacial charge separation in tetraphenyldibenzoperiflanthene:fullerene (DBP:C70) blends are investigated computationally. The analysis is based on a polarization-consistent framework employing screened range-separated hybrid functional in a polarizable continuum model to parametrize Fermi's golden rule rate theory. The model considers the possible transitions within the 25 lowest excited states of a DBP:C70 dyad that are accessible by photoexcitation. The different identified pathways contributing to charge carrier generation include electron and hole transfer and backtransfer, exciton transfer, and internal relaxation steps. The larger density of states of C70 appears to explain the previously observed larger efficiency for charge separation through hole transfer mechanism. We also analyze the validity of the high-temperature and short-time semiclassical approximations of the FGR theory, where both overestimated and underestimated Marcus theory based constants can be affected.

Nanozyme-catalysed CRISPR assay for preamplification-free detection of non-coding RNAs.

CRISPR-based diagnostics enable specific sensing of DNA and RNA biomarkers associated with human diseases. This is achieved through the binding of guide RNAs to a complementary sequence that activates Cas enzymes to cleave reporter molecules. Currently, most CRISPR-based diagnostics rely on target preamplification to reach sufficient sensitivity for clinical applications. This limits quantification capability and adds complexity to the reaction chemistry. Here we show the combination of a CRISPR-Cas-based reaction with a nanozyme-linked immunosorbent assay, which allows for the quantitative and colorimetric readout of Cas13-mediated RNA detection through catalytic metallic nanoparticles at room temperature (CrisprZyme). We demonstrate that CrisprZyme is easily adaptable to a lateral-flow-based readout and different Cas enzymes and enables the sensing of non-coding RNAs including microRNAs, long non-coding RNAs and circular RNAs. We utilize this platform to identify patients with acute myocardial infarction and to monitor cellular differentiation in vitro and in tissue biopsies from prostate cancer patients. We anticipate that CrisprZyme will serve as a universally applicable signal catalyst for CRISPR-based diagnostics, which will expand the spectrum of targets for preamplification-free, quantitative detection.

Activation mechanism of PINK1.

Mutations in the protein kinase PINK1 lead to defects in mitophagy and cause autosomal recessive early onset Parkinson's disease1,2. PINK1 has many unique features that enable it to phosphorylate ubiquitin and the ubiquitin-like domain of Parkin3-9. Structural analysis of PINK1 from diverse insect species10-12 with and without ubiquitin provided snapshots of distinct structural states yet did not explain how PINK1 is activated. Here we elucidate the activation mechanism of PINK1 using crystallography and cryo-electron microscopy (cryo-EM). A crystal structure of unphosphorylated Pediculus humanus corporis (Ph; human body louse) PINK1 resolves an N-terminal helix, revealing the orientation of unphosphorylated yet active PINK1 on the mitochondria. We further provide a cryo-EM structure of a symmetric PhPINK1 dimer trapped during the process of trans-autophosphorylation, as well as a cryo-EM structure of phosphorylated PhPINK1 undergoing a conformational change to an active ubiquitin kinase state. Structures and phosphorylation studies further identify a role for regulatory PINK1 oxidation. Together, our research delineates the complete activation mechanism of PINK1, illuminates how PINK1 interacts with the mitochondrial outer membrane and reveals how PINK1 activity may be modulated by mitochondrial reactive oxygen species.

Modulating the Excited-State Decay Pathways of Cu(I) 4H-Imidazolate Complexes by Excitation Wavelength and Ligand Backbone.

Cu(I) 4H-imidazolato complexes are excellent photosensitizers with broad and intense light absorption properties, based on an earth-abundant metal, and hold great promise as photosensitizers in artificial photosynthesis and for accumulation of redox equivalents. In this study, the excited-state relaxation dynamics of three novel heteroleptic Cu(I) 4H-imidazolato complexes with phenyl, tolyl, and mesityl side groups are systematically investigated by femtosecond and nanosecond time-resolved transient absorption spectroscopy and theoretical methods, complemented by steady-state absorption spectroscopy and (spectro)electrochemistry. After photoexcitation into the metal-to-ligand charge transfer (MLCT) and intraligand charge transfer absorption band, fast (0.6-1 ps) intersystem crossing occurs into the triplet MLCT manifold. The triplet-state population relaxes via the geometrical planarization of the N-aryl rings on the Cu(I) 4H-imidazolato complexes. Depending on the initial Franck-Condon state, the remaining small singlet state population relaxes into two geometrically distinct minima geometries with similar energy, S1/2,relax and S3/4,relax. Subsequent ground-state recovery from S1/2,relax and internal conversion from S3/4,relax to S1/2,relax take place on a 100 ps time scale. The internal conversion can be understood as hole transfer from a dyz-orbital to a dxz-orbital, which is accompanied with the structural reorganization of the coordination environment. Generally, the photophysical processes are determined by the steric hindrance of the side groups on the ligands. And the excited singlet-state pathways are dependent on the excitation wavelength.

Shigella ubiquitin ligase IpaH7.8 targets gasdermin D for degradation to prevent pyroptosis and enable infection.

The pore-forming protein gasdermin D (GSDMD) executes lytic cell death called pyroptosis to eliminate the replicative niche of intracellular pathogens. Evolution favors pathogens that circumvent this host defense mechanism. Here, we show that the Shigella ubiquitin ligase IpaH7.8 functions as an inhibitor of GSDMD. Shigella is an enteroinvasive bacterium that causes hemorrhagic gastroenteritis in primates, but not rodents. IpaH7.8 contributes to species specificity by ubiquitinating human, but not mouse, GSDMD and targeting it for proteasomal degradation. Accordingly, infection of human epithelial cells with IpaH7.8-deficient Shigella flexneri results in increased GSDMD-dependent cell death compared with wild type. Consistent with pyroptosis contributing to murine disease resistance, eliminating GSDMD from NLRC4-deficient mice, which are already sensitized to oral infection with Shigella flexneri, leads to further enhanced bacterial replication and increased disease severity. This work highlights a species-specific pathogen arms race focused on maintenance of host cell viability.

Simulating energy transfer dynamics in the Fenna-Matthews-Olson complex via the modified generalized quantum master equation.

The generalized quantum master equation (GQME) provides a general and formally exact framework for simulating the reduced dynamics of open quantum systems. The recently introduced modified approach to the GQME (M-GQME) corresponds to a specific implementation of the GQME that is geared toward simulating the dynamics of the electronic reduced density matrix in systems governed by an excitonic Hamiltonian. Such a Hamiltonian, which is often used for describing energy and charge transfer dynamics in complex molecular systems, is given in terms of diabatic electronic states that are coupled to each other and correspond to different nuclear Hamiltonians. Within the M-GQME approach, the effect of the nuclear degrees of freedom on the time evolution of the electronic density matrix is fully captured by a memory kernel superoperator, which can be obtained from short-lived (compared to the time scale of energy/charge transfer) projection-free inputs. In this paper, we test the ability of the M-GQME to predict the energy transfer dynamics within a seven-state benchmark model of the Fenna-Matthews-Olson (FMO) complex, with the short-lived projection-free inputs obtained via the Ehrenfest method. The M-GQME with Ehrenfest-based inputs is shown to yield accurate results across a wide parameter range. It is also found to dramatically outperform the direct application of the Ehrenfest method and to provide better-behaved convergence with respect to memory time in comparison to an alternative implementation of the GQME approach previously applied to the same FMO model.

Cyanide Bridged Platinum-Iron Complexes as Cisplatin Prodrug Systems: Design and Computational Study.

The potential role of cyanide-bridged platinum-iron complexes as an anti-cancer Pt(IV) prodrug is studied. We present design principles of a dual-function prodrug that can upon reduction dissociate and release concurrently six cisplatin units and a ferricyanide anion per prodrug unit. The prodrug molecule is a unique complex of hepta metal centers consisting of a ferricyanide core with six Pt(IV) centers each bonded to the Fe(III) core through a cyano ligand. The functionality of the prodrug is addressed through density functional theory (DFT) calculations.

Integration of innate immune signalling by caspase-8 cleavage of N4BP1.

Mutations in the death receptor FAS1,2 or its ligand FASL3 cause autoimmune lymphoproliferative syndrome, whereas mutations in caspase-8 or its adaptor FADD-which mediate cell death downstream of FAS and FASL-cause severe immunodeficiency in addition to autoimmune lymphoproliferative syndrome4-6. Mouse models have corroborated a role for FADD-caspase-8 in promoting inflammatory responses7-12, but the mechanisms that underlie immunodeficiency remain undefined. Here we identify NEDD4-binding protein 1 (N4BP1) as a suppressor of cytokine production that is cleaved and inactivated by caspase-8. N4BP1 deletion in mice increased the production of select cytokines upon stimulation of the Toll-like receptor (TLR)1-TLR2 heterodimer (referred to herein as TLR1/2), TLR7 or TLR9, but not upon engagement of TLR3 or TLR4. N4BP1 did not suppress TLR3 or TLR4 responses in wild-type macrophages, owing to TRIF- and caspase-8-dependent cleavage of N4BP1. Notably, the impaired production of cytokines in response to TLR3 and TLR4 stimulation of caspase-8-deficient macrophages13 was largely rescued by co-deletion of N4BP1. Thus, the persistence of intact N4BP1 in caspase-8-deficient macrophages impairs their ability to mount robust cytokine responses. Tumour necrosis factor (TNF), like TLR3 or TLR4 agonists, also induced caspase-8-dependent cleavage of N4BP1, thereby licensing TRIF-independent TLRs to produce higher levels of inflammatory cytokines. Collectively, our results identify N4BP1 as a potent suppressor of cytokine responses; reveal N4BP1 cleavage by caspase-8 as a point of signal integration during inflammation; and offer an explanation for immunodeficiency caused by mutations of FADD and caspase-8.

Identification and characterization of diverse OTU deubiquitinases in bacteria.

Manipulation of host ubiquitin signaling is becoming an increasingly apparent evolutionary strategy among bacterial and viral pathogens. By removing host ubiquitin signals, for example, invading pathogens can inactivate immune response pathways and evade detection. The ovarian tumor (OTU) family of deubiquitinases regulates diverse ubiquitin signals in humans. Viral pathogens have also extensively co-opted the OTU fold to subvert host signaling, but the extent to which bacteria utilize the OTU fold was unknown. We have predicted and validated a set of OTU deubiquitinases encoded by several classes of pathogenic bacteria. Biochemical assays highlight the ubiquitin and polyubiquitin linkage specificities of these bacterial deubiquitinases. By determining the ubiquitin-bound structures of two examples, we demonstrate the novel strategies that have evolved to both thread an OTU fold and recognize a ubiquitin substrate. With these new examples, we perform the first cross-kingdom structural analysis of the OTU fold that highlights commonalities among distantly related OTU deubiquitinases.

On the Role of the Special Pair in Photosystems as a Charge Transfer Rectifier.

The special pair, a bacteriochlorophyll a (BChl) dimer found at the core of bacterial reaction centers, is known to play a key role in the functionality of photosystems as a precursor to the photosynthesis process. In this paper, we analyze the inherent affinity of the special pair to rectify the intrapair photo-induced charge transfer (CT). In particular, we show that the molecular environment affects the nuclear geometry, resulting in symmetry breaking between the two possible intrapair CT processes. To this end, we study the relationships of the intrapair CT and the molecular geometry with respect to the effective dielectric constant provided by the molecular environment. We identify the special pair structural feature that breaks the symmetry between the two molecules, leading to CT rectification. Excited state energies, oscillator strengths, and electronic coupling values are obtained via time-dependent density functional theory, employing a recently developed framework based on a screened range-separated hybrid functional within a polarizable continuum model (SRSH-PCM). We analyze the rectification capability of the special pair by calculating the CT rates using a first-principles-based Fermi's golden rule approach.

Efficient Charge Generation via Hole Transfer in Dilute Organic Donor-Fullerene Blends.

Efficient organic photovoltaics (OPVs) require broadband charge photogeneration with near-unity quantum yield. This can only be achieved by exploiting all pathways that generate charge. Electron transfer from organic donors to acceptors has been well-studied and is considered the primary path to charge photogeneration in OPVs. In contrast, much less is known about the hole transfer pathway. Here we study charge photogeneration in an archetypal system comprising tetraphenyldibenzoperiflanthene:C70 blends using our recently developed multispectral two-dimensional electronic spectroscopy (M-2DES), supported by time-dependent density functional theory and fully quantum-mechanical Fermi's golden rule rate calculations. Our approach identifies in real time two rapid charge transfer pathways that are confirmed through computational analysis. Surprisingly, we find that both electron and hole transfer occur with comparable rates and efficiencies, facilitated by donor-acceptor electronic interactions. Our results highlight the importance of the hole transfer pathway for optimizing the efficiency of OPV devices employing small-molecule heterojunctions.

Vibronic structure of photosynthetic pigments probed by polarized two-dimensional electronic spectroscopy and ab initio calculations.

Bacteriochlorophyll a (Bchl a) and chlorophyll a (Chl a) play important roles as light absorbers in photosynthetic antennae and participate in the initial charge-separation steps in photosynthetic reaction centers. Despite decades of study, questions remain about the interplay of electronic and vibrational states within the Q-band and its effect on the photoexcited dynamics. Here we report results of polarized two-dimensional electronic spectroscopic measurements, performed on penta-coordinated Bchl a and Chl a and their interpretation based on state-of-the-art time-dependent density functional theory calculations and vibrational mode analysis for spectral shapes. We find that the Q-band of Bchl a is comprised of two independent bands, that are assigned following the Gouterman model to Q x and Q y states with orthogonal transition dipole moments. However, we measure the angle to be ∼75°, a finding that is confirmed by ab initio calculations. The internal conversion rate constant from Q x to Q y is found to be 11 ps-1. Unlike Bchl a, the Q-band of Chl a contains three distinct peaks with different polarizations. Ab initio calculations trace these features back to a spectral overlap between two electronic transitions and their vibrational replicas. The smaller energy gap and the mixing of vibronic states result in faster internal conversion rate constants of 38-50 ps-1. We analyze the spectra of penta-coordinated Bchl a and Chl a to highlight the interplay between low-lying vibronic states and their relationship to photoinduced relaxation. Our findings shed new light on the photoexcited dynamics in photosynthetic systems where these chromophores are primary pigments.

Explaining Spectral Asymmetries and Excitonic Characters of the Core Pigment Pairs in the Bacterial Reaction Center Using a Screened Range-Separated Hybrid Functional.

Spectral peaks of the special pair (P) and adjacent pigments in the bacterial reaction center (BRC) are investigated computationally. We employ a novel framework based on a polarization-consistent treatment of the dielectric environment, combining the polarizable continuum model (PCM) with time-dependent screened range-separated hybrid (SRSH) density functional theory. Our calculations quantitatively reproduce recently measured spectral peak splits between P excitonic states and spectral asymmetries within the pairs of excited states of the adjacent bacteriochlorophyll a (BChl) and bacteriopheophytin a (BPhe) pigments. For the special pair, a splitting energy between the absorptive state and a blue-shifted semidark state of 0.07 eV is found in close agreement with the measured value. The spectral asymmetries within the pseudosymmetric pairs of BChl and BPhe pigments are interpreted to result from locally different effective dielectric environments in the A and the B branch, where the latter are exposed to a lesser polarizing environment. We base our analysis on X-ray-resolved structures and where the effect of neighboring pigments on the electronic structure is addressed through an effective dielectric environment. We show that the spectral trends are only reproduced using a polarization-consistent framework based on a screened range-separated hybrid functional, whereas B3LYP-PCM energies fail to provide the correct trends.

A novel approach for assessing effects of ship traffic on distributions and movements of seabirds.

Marine habitats are nowadays strongly affected by human activities, while for many species the consequences of these impacts are still unclear. The red-throated diver (Gavia stellata) has been reported to be sensitive to ship traffic and other anthropogenic pressures and is consequently of high conservation concern. We studied red-throated divers in the German Bight (North Sea) using satellite telemetry and digital aerial surveys with the aim of assessing effects of ship traffic on the distribution and movements of this species during the non-breeding season. Data from the automatic identification system of ships (AIS) were intersected with bird data and allowed detailed spatial and temporal analyses. During the study period, ship traffic was present throughout the main distribution area of divers. Depending on impact radius, only small areas existed in which ship traffic was present on less than 20% of the days. Ship traffic was dominated by fishing vessels and cargo ships, but also wind farm-related ships were frequently recorded. Red-throated divers were more abundant in areas with no or little concurrent ship traffic. Analysis of aerial survey data revealed strong effects of ship speed on divers: in areas with vessels sailing at high speed only a slow resettlement of the area was observed after the disturbance, while in areas with vessels sailing at medium speed the resettlement was more rapid during the observed time period of 7 hours. Data from satellite-tracking of divers suggest that large relocation distances of individuals are related to disturbance by ships which often trigger birds to take flight. Effective measures to reduce disturbance could include channeled traffic in sensitive areas, as well as speed limits for ships traveling within the protected marine area.

Combining the mapping Hamiltonian linearized semiclassical approach with the generalized quantum master equation to simulate electronically nonadiabatic molecular dynamics.

The generalized quantum master equation (GQME) provides a powerful framework for simulating electronically nonadiabatic molecular dynamics. Within this framework, the effect of the nuclear degrees of freedom on the time evolution of the electronic reduced density matrix is fully captured by a memory kernel superoperator. In this paper, we consider two different procedures for calculating the memory kernel of the GQME from projection-free inputs obtained via the combination of the mapping Hamiltonian (MH) approach and the linearized semiclassical (LSC) approximation. The accuracy and feasibility of the two procedures are demonstrated on the spin-boson model. We find that although simulating the electronic dynamics by direct application of the two LSC-based procedures leads to qualitatively different results that become increasingly less accurate with increasing time, restricting their use to calculating the memory kernel leads to an accurate description of the electronic dynamics. Comparison with a previously proposed procedure for calculating the memory kernel via the Ehrenfest method reveals that MH/LSC methods produce memory kernels that are better behaved at long times and lead to more accurate electronic dynamics.