Opportunities and Challenges for Machine Learning-Assisted Enzyme Engineering.

Enzymes can be engineered at the level of their amino acid sequences to optimize key properties such as expression, stability, substrate range, and catalytic efficiency-or even to unlock new catalytic activities not found in nature. Because the search space of possible proteins is vast, enzyme engineering usually involves discovering an enzyme starting point that has some level of the desired activity followed by directed evolution to improve its "fitness" for a desired application. Recently, machine learning (ML) has emerged as a powerful tool to complement this empirical process. ML models can contribute to (1) starting point discovery by functional annotation of known protein sequences or generating novel protein sequences with desired functions and (2) navigating protein fitness landscapes for fitness optimization by learning mappings between protein sequences and their associated fitness values. In this Outlook, we explain how ML complements enzyme engineering and discuss its future potential to unlock improved engineering outcomes.

DeCOIL: Optimization of Degenerate Codon Libraries for Machine Learning-Assisted Protein Engineering.

With advances in machine learning (ML)-assisted protein engineering, models based on data, biophysics, and natural evolution are being used to propose informed libraries of protein variants to explore. Synthesizing these libraries for experimental screens is a major bottleneck, as the cost of obtaining large numbers of exact gene sequences is often prohibitive. Degenerate codon (DC) libraries are a cost-effective alternative for generating combinatorial mutagenesis libraries where mutations are targeted to a handful of amino acid sites. However, existing computational methods to optimize DC libraries to include desired protein variants are not well suited to design libraries for ML-assisted protein engineering. To address these drawbacks, we present DEgenerate Codon Optimization for Informed Libraries (DeCOIL), a generalized method that directly optimizes DC libraries to be useful for protein engineering: to sample protein variants that are likely to have both high fitness and high diversity in the sequence search space. Using computational simulations and wet-lab experiments, we demonstrate that DeCOIL is effective across two specific case studies, with the potential to be applied to many other use cases. DeCOIL offers several advantages over existing methods, as it is direct, easy to use, generalizable, and scalable. With accompanying software (https://github.com/jsunn-y/DeCOIL), DeCOIL can be readily implemented to generate desired informed libraries.

Mismatch negativity predicts initial auditory-based targeted cognitive training performance in a heterogeneous population across psychiatric disorders.

Auditory-based targeted cognitive training (ATCT) programs are emerging pro-cognitive therapeutic interventions which aim to improve auditory processing to attenuate cognitive impairment in a "bottom up" manner. Biomarkers of early auditory information processing (EAIP) like mismatch negativity (MMN) and P3a have been used successfully to predict gains from a full 40 h course of ATCT in schizophrenia (SZ). Here we investigated the ability of EAIP biomarkers to predict ATCT performance in a group of subjects (n = 26) across SZ, MDD, PTSD and GAD diagnoses. Cognition was assessed via the MATRICS Consensus Cognitive Battery (MCCB) and MMN/P3a were collected prior to completing 1 h of "Sound Sweeps," a representative ATCT exercise. Baseline and final performance over the first two levels of cognitive training served as the primary dependent variables. Groups had similar MMN, but the SZ group had attenuated P3a. MMN and MCCB cognitive domain t-scores, but not P3a, were strongly correlated with most ATCT performance measures, and explained up to 61% of variance in ATCT performance. Diagnosis was not a significant predictor for ATCT performance. These data suggest that MMN can predict ATCT performance in heterogeneous neuropsychiatric populations and should be considered in ATCT studies across diagnostically diverse cohorts.

A meta-analysis of social cognitive deficits in schizophrenia: Does world region matter?

Individuals with schizophrenia (SCZs) demonstrate social cognitive (SC) deficits across a variety of social cognitive tasks: affect recognition, social perception, Theory-of-Mind (ToM), and attributional style. These tasks all use socially-relevant words, social scenarios, clothing, faces and voices that may make their application to different cultural contexts problematic. However a paucity of research has investigated the cross-cultural validity of SC measures. We meta-analyzed the literature investigating differences in social cognitive skills SCZs and matched healthy controls (NCs) across different regions of the world using a group of expert-selected, standardized measures of social cognition. Studies of SC in SCZ using these measures published between January 1980 and August 2020 were evaluated. Data were extracted independently by 3 reviewers with excellent reliability; 156 unique studies of 10,235 SCZs and 9924 NCs across 34 countries were identified. Random effects models revealed SCZs demonstrated poorer performance in all domains of SC including emotion processing (g = -0.770), social perception (g = -0.880), ToM (g = -1.090), attributional style (hostility: g = -0.715, aggression: g = -0.209, blame: g = -0.322), as well as a measure of emotion regulation (g = -0.867). Hostile attributional style was more pronounced in European and North American samples (g = 1.054 and g = -0.605, respectively) compared to Asian samples (g = -0.284). Our results revealed that SCZs performed mildly-severely worse than HCs in all domains of SC. With the exception of hostile attributional style, the magnitude of deficits in social cognition was consistent across the globe.

Two-dimensional multiplexed assay for rapid and deep SARS-CoV-2 serology profiling and for machine learning prediction of neutralization capacity.

Antibody responses serve as the primary protection against SARS-CoV-2 infection through neutralization of viral entry into cells. We have developed a two-dimensional multiplex bead binding assay (2D-MBBA) that quantifies multiple antibody isotypes against multiple antigens from a single measurement. Here, we applied our assay to profile IgG, IgM and IgA levels against the spike antigen, its receptor-binding domain and natural and designed mutants. Machine learning algorithms trained on the 2D-MBBA data substantially improve the prediction of neutralization capacity against the authentic SARS-CoV-2 virus of serum samples of convalescent patients. The algorithms also helped identify a set of antibody isotype-antigen datasets that contributed to the prediction, which included those targeting regions outside the receptor-binding interface of the spike protein. We applied the assay to profile samples from vaccinated, immune-compromised patients, which revealed differences in the antibody profiles between convalescent and vaccinated samples. Our approach can rapidly provide deep antibody profiles and neutralization prediction from essentially a drop of blood without the need of BSL-3 access and provides insights into the nature of neutralizing antibodies. It may be further developed for evaluating neutralizing capacity for new variants and future pathogens.

Iterative screening methodology enables isolation of strains with improved properties for a FACS-based screen and increased L-DOPA production.

Optimizing microbial hosts for the large-scale production of valuable metabolites often requires multiple mutations and modifications to the host's genome. We describe a three-round screen for increased L-DOPA production in S. cerevisiae using FACS enrichment of an enzyme-coupled biosensor for L-DOPA. Multiple rounds of screening were enabled by a single build of a barcoded in vitro transposon-mediated disruption library. New background strains for screening were built for each iteration using results from previous iterations. The same in vitro transposon-mediated disruption library was integrated by homologous recombination into new background strains in each round of screening. Compared with creating new transposon insertions in each round, this method takes less time and saves the cost of additional sequencing to characterize transposon insertion sites. In the first two rounds of screening, we identified deletions that improved biosensor compartmentalization and, consequently, improved our ability to screen for L-DOPA production. In a final round, we discovered that deletion of heme oxygenase (HMX1) increases total heme concentration and increases L-DOPA production, using dopamine measurement as a proxy. We further demonstrated that deleting HMX1 may represent a general strategy for P450 function improvement by improving activity of a second P450 enzyme, BM3, which performs a distinct reaction.

The impact of glycaemic variability on wound healing in the diabetic foot - A retrospective study of new ulcers presenting to a specialist multidisciplinary foot clinic.

AIMS: Glycaemic variability - the visit-to-visit variation in HbA1c - plays a possible role in the development of micro and macrovascular disease in patients with diabetes. Whether HbA1c variability is a factor determining wound healing in diabetic foot ulcers remains unknown. We aimed to determine whether HbA1c variability is associated with foot ulcer healing time. METHODS: A retrospective analysis of patients presenting to our specialist multidisciplinary foot clinic between July 2013 and March 2015, with at least three HbA1c measurements within five years of presentation and more than two follow-up reviews. HbA1c variation was measured by magnitude of standard deviation. RESULTS: 629 new referrals were seen between July 2013 and March 2015. Of these, 172 patients had their number of days to healing recorded and sufficient numbers of HbA1c values to determine variability. The overall geometric mean days to heal was 91.1 days (SD 80.8-102.7). In the low HbA1c variability group the geometric mean days to heal was 78.0 days (60.2-101.2) vs 126.9 days (102.0-158.0) in the high Hb1Ac variability group (p = .032). Those with low HbA1c (<58 mmol/mol) and low variability healed faster than those with high HbA1c and high variability (73.5 days [59.5-90.8] vs 111.0 days [92.0-134.0], p = .007). Additionally, our results show that time to healing is more dependent on the mean HbA1c than the variability in HbA1c (p = .007). CONCLUSIONS/INTERPRETATION: Our data suggest that there was a significant association between HbA1c variability and healing time in diabetic foot ulcers.

Effectiveness of interventions to directly support food and drink intake in people with dementia: systematic review and meta-analysis.

BACKGROUND: Eating and drinking difficulties are recognised sources of ill health in people with dementia. In the EDWINA (Eating and Drinking Well IN dementiA) systematic review we aimed to assess effectiveness of interventions to directly improve, maintain or facilitate oral food and drink intake, nutrition and hydration status, in people with cognitive impairment or dementia (across all settings, levels of care and support, types and degrees of dementia). Interventions included oral nutrition supplementation, food modification, dysphagia management, eating assistance and supporting the social element of eating and drinking. METHODS: We comprehensively searched 13 databases for relevant intervention studies. The review was conducted with service user input in accordance with Cochrane Collaboration's guidelines. We duplicated assessment of inclusion, data extraction, and validity assessment, tabulating data, carrying out random effects meta-analysis and narrative synthesis. RESULTS: Forty-three controlled interventions were included, disappointingly none were judged at low risk of bias. Oral nutritional supplementation studies suggested small positive short term but unclear long term effects on nutritional status. Food modification or dysphagia management studies were smaller and of low quality, providing little evidence of an improved nutritional status. Eating assistance studies provided inconsistent evidence, but studies with a strong social element around eating/drinking, although small and of low quality provided consistent suggestion of improvements in aspects of quality of life. There were few data to address stakeholders' questions. CONCLUSIONS: We found no definitive evidence on effectiveness, or lack of effectiveness, of specific interventions but studies were small and short term. People with cognitive impairment and their carers have to tackle eating problems despite this lack of evidence, so promising interventions are listed. The need remains for high quality trials tailored for people with cognitive impairment assessing robust outcomes. SYSTEMATIC REVIEW REGISTRATION: The systematic review protocol was registered (CRD42014007611) and is published, with the full MEDLINE search strategy, on Prospero.

Testing the effect of a targeted intervention on nurses' compliance with "best practice" mechanical venous thromboembolism prevention.

The aim of this study was to examine whether educational outreach visits improve nurses' compliance with applying best practice mechanical venous thromboembolism prophylaxis. The design was as a pretest/posttest study with a 7-week follow-up. It was conducted in a mixed medical/surgical unit in a 250-bed private hospital in Sydney, Australia. The target population was 25 medical/surgical nurses in educational outreach visits (EOVs). The main outcome measures included change in percentage between baseline and endpoint of eligible patients receiving mechanical VTE prophylaxis and all patients having VTE risk documented in their medication charts, as well as nurses' feedback on how supportive and useful they found EOVs. The results showed an overall, but not significant increase (p = 0.201) in the percentage of patients who received mechanical VTE prophylaxis (59.4% baseline to 75% endpoint). There was a significant increase in the percentage of patients having VTE risk status documented in the medication chart (0%-28%) (p = 0.002). Improvements in compliance were more likely for surgical than medical patients (95% and 35%, respectively) and risk documentation (47% and 6%, respectively). Most nurses reported that the EOVs supported them in implementing best practice VTE mechanical prophylaxis. Researchers conclude that improvements in compliance with best practice VTE prevention can be achieved using EOVs which were easily conducted and well-received in a busy unit setting. More work is needed to increase the compliance rate with medical patients.