grünifai: interactive multiparameter optimization of molecules in a continuous vector space.

SUMMARY: Optimizing small molecules in a drug discovery project is a notoriously difficult task as multiple molecular properties have to be considered and balanced at the same time. In this work, we present our novel interactive in silico compound optimization platform termed grünifai to support the ideation of the next generation of compounds under the constraints of a multiparameter objective. grünifai integrates adjustable in silico models, a continuous representation of the chemical space, a scalable particle swarm optimization algorithm and the possibility to actively steer the compound optimization through providing feedback on generated intermediate structures. AVAILABILITY AND IMPLEMENTATION: Source code and documentation are freely available under an MIT license and are openly available on GitHub (https://github.com/jrwnter/gruenifai). The backend, including the optimization method and distribution on multiple GPU nodes is written in Python 3. The frontend is written in ReactJS.

Unsupervised Representation Learning for Proteochemometric Modeling.

In silico protein-ligand binding prediction is an ongoing area of research in computational chemistry and machine learning based drug discovery, as an accurate predictive model could greatly reduce the time and resources necessary for the detection and prioritization of possible drug candidates. Proteochemometric modeling (PCM) attempts to create an accurate model of the protein-ligand interaction space by combining explicit protein and ligand descriptors. This requires the creation of information-rich, uniform and computer interpretable representations of proteins and ligands. Previous studies in PCM modeling rely on pre-defined, handcrafted feature extraction methods, and many methods use protein descriptors that require alignment or are otherwise specific to a particular group of related proteins. However, recent advances in representation learning have shown that unsupervised machine learning can be used to generate embeddings that outperform complex, human-engineered representations. Several different embedding methods for proteins and molecules have been developed based on various language-modeling methods. Here, we demonstrate the utility of these unsupervised representations and compare three protein embeddings and two compound embeddings in a fair manner. We evaluate performance on various splits of a benchmark dataset, as well as on an internal dataset of protein-ligand binding activities and find that unsupervised-learned representations significantly outperform handcrafted representations.

Spin States Protected from Intrinsic Electron-Phonon Coupling Reaching 100 ns Lifetime at Room Temperature in MoSe2.

We present time-resolved Kerr rotation measurements, showing spin lifetimes of over 100 ns at room temperature in monolayer MoSe2. These long lifetimes are accompanied by an intriguing temperature-dependence of the Kerr amplitude, which increases with temperature up to 50 K and then abruptly switches sign. Using ab initio simulations, we explain the latter behavior in terms of the intrinsic electron-phonon coupling and the activation of transitions to secondary valleys. The phonon-assisted scattering of the photoexcited electron-hole pairs prepares a valley spin polarization within the first few ps after laser excitation. The sign of the total valley magnetization, and thus the Kerr amplitude, switches as a function of temperature, as conduction and valence band states exhibit different phonon-mediated intervalley scattering rates. However, the electron-phonon scattering on the ps time scale does not provide an explanation for the long spin lifetimes. Hence, we deduce that the initial spin polarization must be transferred into spin states, which are protected from the intrinsic electron-phonon coupling, and are most likely resident charge carriers, which are not part of the itinerant valence or conduction band states.

Promoting improved social support and quality of life with the CenteringPregnancy® group model of prenatal care.

This prospective cohort study compared women participating in CenteringPregnancy® group prenatal care (N = 120) with those in standard individual care (N = 221) to determine if participation in Centering was associated with improvements in perceived social support and quality of life, with concomitant decreases in screens of postpartum depression and improvements in breastfeeding rates. Participants completed surveys at the onset of prenatal care, in the late third trimester and in the postpartum period. Centering participants had higher scores of perceived social support from friends after participating in group care (p < 0.05) with associated improvements in quality of life in the psychological and relational domains (p < 0.05) compared to standard care participants who showed higher scores of perceived support from family (p < 0.05) but did not show concomitant improvements in quality of life. This did not translate to any significant difference in scores on postpartum depression screens but was associated with improvements in breastfeeding continuation rates among Centering participants in the postpartum period. This study indicates that Centering care is associated with improved perceptions of peer social support with associated improvements in quality of life and higher rates of continued breastfeeding.

Mechanical constraints imposed by 3D cellular geometry and arrangement modulate growth patterns in the Arabidopsis embryo.

Morphogenesis occurs in 3D space over time and is guided by coordinated gene expression programs. Here we use postembryonic development in Arabidopsis plants to investigate the genetic control of growth. We demonstrate that gene expression driving the production of the growth-stimulating hormone gibberellic acid and downstream growth factors is first induced within the radicle tip of the embryo. The center of cell expansion is, however, spatially displaced from the center of gene expression. Because the rapidly growing cells have very different geometry from that of those at the tip, we hypothesized that mechanical factors may contribute to this growth displacement. To this end we developed 3D finite-element method models of growing custom-designed digital embryos at cellular resolution. We used this framework to conceptualize how cell size, shape, and topology influence tissue growth and to explore the interplay of geometrical and genetic inputs into growth distribution. Our simulations showed that mechanical constraints are sufficient to explain the disconnect between the experimentally observed spatiotemporal patterns of gene expression and early postembryonic growth. The center of cell expansion is the position where genetic and mechanical facilitators of growth converge. We have thus uncovered a mechanism whereby 3D cellular geometry helps direct where genetically specified growth takes place.

Identification of mutations in CUL7 in 3-M syndrome.

Intrauterine growth retardation is caused by maternal, fetal or placental factors that result in impaired endovascular trophoblast invasion and reduced placental perfusion. Although various causes of intrauterine growth retardation have been identified, most cases remain unexplained. Studying 29 families with 3-M syndrome (OMIM 273750), an autosomal recessive condition characterized by severe pre- and postnatal growth retardation, we first mapped the underlying gene to chromosome 6p21.1 and then identified 25 distinct mutations in the gene cullin 7 (CUL7). CUL7 assembles an E3 ubiquitin ligase complex containing Skp1, Fbx29 (also called Fbw8) and ROC1 and promotes ubiquitination. Using deletion analysis, we found that CUL7 uses its central region to interact with the Skp1-Fbx29 heterodimer. Functional studies indicated that the 3-M-associated CUL7 nonsense and missense mutations R1445X and H1464P, respectively, render CUL7 deficient in recruiting ROC1. These results suggest that impaired ubiquitination may have a role in the pathogenesis of intrauterine growth retardation in humans.

A community effort in SARS-CoV-2 drug discovery.

The COVID-19 pandemic continues to pose a substantial threat to human lives and is likely to do so for years to come. Despite the availability of vaccines, searching for efficient small-molecule drugs that are widely available, including in low- and middle-income countries, is an ongoing challenge. In this work, we report the results of an open science community effort, the "Billion molecules against COVID-19 challenge", to identify small-molecule inhibitors against SARS-CoV-2 or relevant human receptors. Participating teams used a wide variety of computational methods to screen a minimum of 1 billion virtual molecules against 6 protein targets. Overall, 31 teams participated, and they suggested a total of 639,024 molecules, which were subsequently ranked to find 'consensus compounds'. The organizing team coordinated with various contract research organizations (CROs) and collaborating institutions to synthesize and test 878 compounds for biological activity against proteases (Nsp5, Nsp3, TMPRSS2), nucleocapsid N, RdRP (only the Nsp12 domain), and (alpha) spike protein S. Overall, 27 compounds with weak inhibition/binding were experimentally identified by binding-, cleavage-, and/or viral suppression assays and are presented here. Open science approaches such as the one presented here contribute to the knowledge base of future drug discovery efforts in finding better SARS-CoV-2 treatments.

Fraser syndrome and mouse blebbed phenotype caused by mutations in FRAS1/Fras1 encoding a putative extracellular matrix protein.

Fraser syndrome (OMIM 219000) is a multisystem malformation usually comprising cryptophthalmos, syndactyly and renal defects. Here we report autozygosity mapping and show that the locus FS1 at chromosome 4q21 is associated with Fraser syndrome, although the condition is genetically heterogeneous. Mutation analysis identified five frameshift mutations in FRAS1, which encodes one member of a family of novel proteins related to an extracellular matrix (ECM) blastocoelar protein found in sea urchin. The FRAS1 protein contains a series of N-terminal cysteine-rich repeat motifs previously implicated in BMP metabolism, suggesting that it has a role in both structure and signal propagation in the ECM. It has been speculated that Fraser syndrome is a human equivalent of the blebbed phenotype in the mouse, which has been associated with mutations in at least five loci including bl. As mapping data were consistent with homology of FRAS1 and bl, we screened DNA from bl/bl mice and identified a premature termination of mouse Fras1. Thus, the bl mouse is a model for Fraser syndrome in humans, a disorder caused by disrupted epithelial integrity in utero.

Img2Mol - accurate SMILES recognition from molecular graphical depictions.

The automatic recognition of the molecular content of a molecule's graphical depiction is an extremely challenging problem that remains largely unsolved despite decades of research. Recent advances in neural machine translation enable the auto-encoding of molecular structures in a continuous vector space of fixed size (latent representation) with low reconstruction errors. In this paper, we present a fast and accurate model combining deep convolutional neural network learning from molecule depictions and a pre-trained decoder that translates the latent representation into the SMILES representation of the molecules. This combination allows us to precisely infer a molecular structure from an image. Our rigorous evaluation shows that Img2Mol is able to correctly translate up to 88% of the molecular depictions into their SMILES representation. A pretrained version of Img2Mol is made publicly available on GitHub for non-commercial users.

Neuraldecipher - reverse-engineering extended-connectivity fingerprints (ECFPs) to their molecular structures.

Protecting molecular structures from disclosure against external parties is of great relevance for industrial and private associations, such as pharmaceutical companies. Within the framework of external collaborations, it is common to exchange datasets by encoding the molecular structures into descriptors. Molecular fingerprints such as the extended-connectivity fingerprints (ECFPs) are frequently used for such an exchange, because they typically perform well on quantitative structure-activity relationship tasks. ECFPs are often considered to be non-invertible due to the way they are computed. In this paper, we present a fast reverse-engineering method to deduce the molecular structure given revealed ECFPs. Our method includes the Neuraldecipher, a neural network model that predicts a compact vector representation of compounds, given ECFPs. We then utilize another pre-trained model to retrieve the molecular structure as SMILES representation. We demonstrate that our method is able to reconstruct molecular structures to some extent, and improves, when ECFPs with larger fingerprint sizes are revealed. For example, given ECFP count vectors of length 4096, we are able to correctly deduce up to 69% of molecular structures on a validation set (112 K unique samples) with our method.

Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations.

There has been a recent surge of interest in using machine learning across chemical space in order to predict properties of molecules or design molecules and materials with the desired properties. Most of this work relies on defining clever feature representations, in which the chemical graph structure is encoded in a uniform way such that predictions across chemical space can be made. In this work, we propose to exploit the powerful ability of deep neural networks to learn a feature representation from low-level encodings of a huge corpus of chemical structures. Our model borrows ideas from neural machine translation: it translates between two semantically equivalent but syntactically different representations of molecular structures, compressing the meaningful information both representations have in common in a low-dimensional representation vector. Once the model is trained, this representation can be extracted for any new molecule and utilized as a descriptor. In fair benchmarks with respect to various human-engineered molecular fingerprints and graph-convolution models, our method shows competitive performance in modelling quantitative structure-activity relationships in all analysed datasets. Additionally, we show that our descriptor significantly outperforms all baseline molecular fingerprints in two ligand-based virtual screening tasks. Overall, our descriptors show the most consistent performances in all experiments. The continuity of the descriptor space and the existence of the decoder that permits deducing a chemical structure from an embedding vector allow for exploration of the space and open up new opportunities for compound optimization and idea generation.

Efficient multi-objective molecular optimization in a continuous latent space.

One of the main challenges in small molecule drug discovery is finding novel chemical compounds with desirable properties. In this work, we propose a novel method that combines in silico prediction of molecular properties such as biological activity or pharmacokinetics with an in silico optimization algorithm, namely Particle Swarm Optimization. Our method takes a starting compound as input and proposes new molecules with more desirable (predicted) properties. It navigates a machine-learned continuous representation of a drug-like chemical space guided by a defined objective function. The objective function combines multiple in silico prediction models, defined desirability ranges and substructure constraints. We demonstrate that our proposed method is able to consistently find more desirable molecules for the studied tasks in relatively short time. We hope that our method can support medicinal chemists in accelerating and improving the lead optimization process.

A 3-Year Study of Resident Reaction to 2011 ACGME Work Hour Rules in a Family Medicine Residency.

INTRODUCTION: 2011 Accreditation Council for Graduate Medical Education (ACGME) work hour rules prompted concerns regarding potential negative impacts on patient care and resident education. We were interested in resident reaction to call restructuring and night oat (NF) in a family medicine residency over 3 years following implementation of the 2011 rules. METHODS: We conducted structured interviews of residents from 2011-2012 through 2013-2014. Interviews were recorded, transcribed, and analyzed for themes. RESULTS: Fifty-eight interviews were conducted, including 18/18 residents in 2011-2012 (100%), 18/20 residents in 2012-2013 (90%), and 22/22 residents in 2013-2014 (100%). Following introduction of the 24-hour work limit, upper year residents reported significantly less fatigue and improved personal lives, patient care, and educational experience. Reactions to NF varied with length and intensity of the NF rotation; most PGY-1 residents reported increased fatigue, more burnout, and worse personal lives on NF. Most residents felt patient care quality on NF did not differ from non-NF rotations because improved inpatient nighttime continuity mitigated effects of fatigue and increased care transitions. Reactions regarding educational experience on NF were initially negative, but improved over time. CONCLUSIONS: Residents' reactions to 2011 ACGME work hour rules suggest the rules improved resident well-being, except on NF. Negative effects of NF may be minimized by limiting NF rotations to 5 nights/week for 2 consecutive weeks, and 1 month total per academic year.

Resident knowledge acquisition during a block conference series.

OBJECTIVE: This study's objective was to determine whether attendance at lectures in a block conference format improves residents' knowledge. METHODS: Seventeen family medicine residents were tested on the content of 27 lectures delivered in a block conference format over a 6-month period. For each lecture, residents completed a pretest, a short-term posttest, and a long-term posttest (1--3 weeks and 1.5--6 months after each lecture, respectively). RESULTS: Mean short-term posttest scores were 10.3 points higher for lecture attendees than nonattendees. Mean long-term posttest scores did not differ significantly for attendees (62.2) versus nonattendees (60.0). CONCLUSIONS: Attendance at didactic lectures in a block conference format did not improve resident knowledge over the long term. These results lead us to question the value of a block conference format and raise the possibility that resident learning might be better served by maximizing clinical experiences and minimizing time in conferences.

No evidence for submicroscopic 22qter deletions in patients with features suggestive for Angelman syndrome.

Patients with monosomy 22q13.3 --> qter have, in addition to (usually severe) developmental delay, hypotonia, severe expressive language delay leading to absence of speech, pervasive developmental abnormalities, and subtle facial anomalies. Thus far, it has been one of the more common submicroscopic telomere deletions seen in patients with mental retardation. Due to the phenotypic overlap between monosomy 22q13.3 and Angelman syndrome (AS), 44 patients with AS features but without one of the characteristic molecular 15q abnormalities were tested for 22qter deletions. In the study group, 31/44 (70%) were heterozygous for locus D22S163 with probe cMS607 (distance 0.125 Mb from telomere). The remaining 13/44 (30%) patients were heterozygous for one or more of four microsatellite markers centromeric from D22S163 in the 22qter region (distances 1.5-4.3 Mb from telomere). Based on the present study, there is no evidence that patients with an "Angelman-like" phenotype are more likely to have a 22qter deletion than other individuals with mental retardation.

Pseudodominant inheritance of Langer mesomelic dysplasia caused by a SHOX homeobox missense mutation.

We report the clinical and molecular analysis in a consanguineous family in which the skeletal dysplasias Léri-Weill dyschondrosteosis (LWD) and Langer mesomelic dysplasia (LMD) both segregate. A newborn male and his mother, both with Langer mesomelic dysplasia, are described. A homozygous SHOX homeobox point mutation, C517T, was identified by direct sequencing in the proband and his mother. The same mutation was present in the heterozygous state in the proband's father and in the maternal grandmother, both of whom had features of LWD. This C to T transition is predicted to cause an arginine to cysteine amino acid change in a highly conserved region of the recognition helix of the homeodomain, which may reduce the stability of the interaction between the SHOX protein and its target DNA. In addition, the mutation may disrupt a nuclear localization signal in SHOX. This is the first SHOX point mutation identified in a case of LMD, and the first case in which parent to child transmission of LMD has been described.

Working with impaired residents: trials, tribulations, and successes.

Impairment of physicians' ability to practice medicine safely and effectively is relatively common. Chemical dependency, the leading cause of physician impairment, has a lifetime prevalence of approximately 10%-15% among physicians. Statistics from physician health programs indicate that family physicians are overrepresented among impaired physicians. It is therefore important for family practice residencies to monitor for and deal with physician impairment. Over the past 11 years, we have worked with eight impaired residents: five with chemical dependency, two with cognitive impairment, and one with an affective disorder Seven of the eight residents are currently practicing medicine, six in family practice. Based on our experience and the literature, we have developed an algorithm that includes the recognition, intervention, and aftercare of impaired residents. The long-term success of the majority of impaired residents with whom we have worked suggests that the trials and tribulations of working with this potentially difficult group of residents are well worth the effort.