Deep learning models for predicting RNA degradation via dual crowdsourcing.

Medicines based on messenger RNA (mRNA) hold immense potential, as evidenced by their rapid deployment as COVID-19 vaccines. However, worldwide distribution of mRNA molecules has been limited by their thermostability, which is fundamentally limited by the intrinsic instability of RNA molecules to a chemical degradation reaction called in-line hydrolysis. Predicting the degradation of an RNA molecule is a key task in designing more stable RNA-based therapeutics. Here, we describe a crowdsourced machine learning competition ('Stanford OpenVaccine') on Kaggle, involving single-nucleotide resolution measurements on 6,043 diverse 102-130-nucleotide RNA constructs that were themselves solicited through crowdsourcing on the RNA design platform Eterna. The entire experiment was completed in less than 6 months, and 41% of nucleotide-level predictions from the winning model were within experimental error of the ground truth measurement. Furthermore, these models generalized to blindly predicting orthogonal degradation data on much longer mRNA molecules (504-1,588 nucleotides) with improved accuracy compared with previously published models. These results indicate that such models can represent in-line hydrolysis with excellent accuracy, supporting their use for designing stabilized messenger RNAs. The integration of two crowdsourcing platforms, one for dataset creation and another for machine learning, may be fruitful for other urgent problems that demand scientific discovery on rapid timescales.

Deep learning models for predicting RNA degradation via dual crowdsourcing.

Messenger RNA-based medicines hold immense potential, as evidenced by their rapid deployment as COVID-19 vaccines. However, worldwide distribution of mRNA molecules has been limited by their thermostability, which is fundamentally limited by the intrinsic instability of RNA molecules to a chemical degradation reaction called in-line hydrolysis. Predicting the degradation of an RNA molecule is a key task in designing more stable RNA-based therapeutics. Here, we describe a crowdsourced machine learning competition ("Stanford OpenVaccine") on Kaggle, involving single-nucleotide resolution measurements on 6043 102-130-nucleotide diverse RNA constructs that were themselves solicited through crowdsourcing on the RNA design platform Eterna. The entire experiment was completed in less than 6 months, and 41% of nucleotide-level predictions from the winning model were within experimental error of the ground truth measurement. Furthermore, these models generalized to blindly predicting orthogonal degradation data on much longer mRNA molecules (504-1588 nucleotides) with improved accuracy compared to previously published models. Top teams integrated natural language processing architectures and data augmentation techniques with predictions from previous dynamic programming models for RNA secondary structure. These results indicate that such models are capable of representing in-line hydrolysis with excellent accuracy, supporting their use for designing stabilized messenger RNAs. The integration of two crowdsourcing platforms, one for data set creation and another for machine learning, may be fruitful for other urgent problems that demand scientific discovery on rapid timescales.

A community-powered search of machine learning strategy space to find NMR property prediction models.

The rise of machine learning (ML) has created an explosion in the potential strategies for using data to make scientific predictions. For physical scientists wishing to apply ML strategies to a particular domain, it can be difficult to assess in advance what strategy to adopt within a vast space of possibilities. Here we outline the results of an online community-powered effort to swarm search the space of ML strategies and develop algorithms for predicting atomic-pairwise nuclear magnetic resonance (NMR) properties in molecules. Using an open-source dataset, we worked with Kaggle to design and host a 3-month competition which received 47,800 ML model predictions from 2,700 teams in 84 countries. Within 3 weeks, the Kaggle community produced models with comparable accuracy to our best previously published 'in-house' efforts. A meta-ensemble model constructed as a linear combination of the top predictions has a prediction accuracy which exceeds that of any individual model, 7-19x better than our previous state-of-the-art. The results highlight the potential of transformer architectures for predicting quantum mechanical (QM) molecular properties.

Reconstruction of Compton Edges in Plastic Gamma Spectra Using Deep Autoencoder.

Plastic scintillation detectors are widely utilized in radiation measurement because of their unique characteristics. However, they are generally used for counting applications because of the energy broadening effect and the absence of a photo peak in their spectra. To overcome their weaknesses, many studies on pseudo spectroscopy have been reported, but most of them have not been able to directly identify the energy of incident gamma rays. In this paper, we propose a method to reconstruct Compton edges in plastic gamma spectra using an artificial neural network for direct pseudo gamma spectroscopy. Spectra simulated using MCNP 6.2 software were used to generate training and validation sets. Our model was trained to reconstruct Compton edges in plastic gamma spectra. In addition, we aimed for our model to be capable of reconstructing Compton edges even for spectra having poor counting statistics by designing a dataset generation procedure. Minimum reconstructible counts for single isotopes were evaluated with metric of mean averaged percentage error as 650 for 60Co, 2000 for 137Cs, 3050 for 22Na, and 3750 for 133Ba. The performance of our model was verified using the simulated spectra measured by a PVT detector. Although our model was trained using simulation data only, it successfully reconstructed Compton edges even in measured gamma spectra with poor counting statistics.

Finding Hidden Signals in Chemical Sensors Using Deep Learning.

Achieving high signal-to-noise ratio in chemical and biological sensors enables accurate detection of target analytes. Unfortunately, below the limit of detection (LOD), it becomes difficult to detect the presence of small amounts of analytes and extract useful information via any of the conventional methods. In this work, we examine the possibility of extracting "hidden signals" using deep neural network to enhance gas sensing below the LOD region. As a test case system, we conduct experiments for H2 sensing in six different metallic channels (Au, Cu, Mo, Ni, Pt, Pd) and demonstrate that deep neural network can enhance the sensing capabilities for H2 concentration below the LOD. We demonstrate that this technique could be universally used for different types of sensors and target analytes. Our approach can extract new information from the hidden signals, which can be crucial for next-generation chemical sensing applications and analytical chemistry.

Probing gas adsorption in MOFs using an efficient ab initio widom insertion Monte Carlo method.

We propose a novel biased Widom insertion method that can efficiently compute the Henry coefficient, KH , of gas molecules inside porous materials exhibiting strong adsorption sites by employing purely DFT calculations. This is achieved by partitioning the simulation volume into strongly and weakly adsorbing regions and selectively biasing the Widom insertion moves into the former region. We show that only few thousands of single point energy calculations are necessary to achieve accurate statistics compared to many hundreds of thousands or millions of such calculations in conventional random insertions. The methodology is used to compute the Henry coefficient for CO2 , N2 , CH4 , and C2 H2 in M-MOF-74(M = Zn and Mg), yielding good agreement with published experimental data. Our results demonstrate that the DFT binding energy and the heat of adsorption are not accurate enough indicators to rank the guest adsorption properties at the Henry regime. © 2016 Wiley Periodicals, Inc.

Superior Chemical Sensing Performance of Black Phosphorus: Comparison with MoS2 and Graphene.

Superior chemical sensing performance of black phosphorus (BP) is demonstrated by comparison with MoS2 and graphene. Dynamic sensing measurements of multichannel detection show that BP displays highly sensitive, selective, and fast-responsive NO2 sensing performance compared to the other representative 2D sensing materials.

Highly Enhanced Gas Adsorption Properties in Vertically Aligned MoS2 Layers.

In this work, we demonstrate that gas adsorption is significantly higher in edge sites of vertically aligned MoS2 compared to that of the conventional basal plane exposed MoS2 films. To compare the effect of the alignment of MoS2 on the gas adsorption properties, we synthesized three distinct MoS2 films with different alignment directions ((1) horizontally aligned MoS2 (basal plane exposed), (2) mixture of horizontally aligned MoS2 and vertically aligned layers (basal and edge exposed), and (3) vertically aligned MoS2 (edge exposed)) by using rapid sulfurization method of CVD process. Vertically aligned MoS2 film shows about 5-fold enhanced sensitivity to NO2 gas molecules compared to horizontally aligned MoS2 film. Vertically aligned MoS2 has superior resistance variation compared to horizontally aligned MoS2 even with same surface area exposed to identical concentration of gas molecules. We found that electrical response to target gas molecules correlates directly with the density of the exposed edge sites of MoS2 due to high adsorption of gas molecules onto edge sites of vertically aligned MoS2. Density functional theory (DFT) calculations corroborate the experimental results as stronger NO2 binding energies are computed for multiple configurations near the edge sites of MoS2, which verifies that electrical response to target gas molecules (NO2) correlates directly with the density of the exposed edge sites of MoS2 due to high adsorption of gas molecules onto edge sites of vertically aligned MoS2. We believe that this observation extends to other 2D TMD materials as well as MoS2 and can be applied to significantly enhance the gas sensor performance in these materials.

The adipokine Retnla modulates cholesterol homeostasis in hyperlipidemic mice.

Hyperlipidemia is a well-recognized risk factor for atherosclerosis and can be regulated by adipokines. Expression of the adipokine resistin-like molecule alpha (Retnla) is regulated by food intake; whether Retnla has a role in the pathogenesis of hyperlipidemia and atherosclerosis is unknown. Here we report that Retnla has a cholesterol-lowering effect and protects against atherosclerosis in low-density lipoprotein receptor-deficient mice. On a high-fat diet, Retnla deficiency promotes hypercholesterolaemia and atherosclerosis, whereas Retnla overexpression reverses these effects and improves the serum lipoprotein profile, with decreased cholesterol in the very low-density lipoprotein fraction concomitant with reduced serum apolipoprotein B levels. We show that Retnla upregulates cholesterol-7-α-hydroxylase, a key hepatic enzyme in the cholesterol catabolic pathway, through induction of its transcriptional activator liver receptor homologue-1, leading to increased excretion of cholesterol in the form of bile acids. These findings define Retnla as a novel therapeutic target for treating hypercholesterolaemia and atherosclerosis.

Evaluation of VCAM-1 antibodies as therapeutic agent for atherosclerosis in apolipoprotein E-deficient mice.

OBJECTIVE: Blocking agents targeting cell adhesion molecules have been developed to prevent cardiovascular diseases such as atherosclerosis, whereas relatively little attention has been paid to the therapeutic potential of vascular cell adhesion molecule (VCAM)-1 as an inflammatory disease target. Two novel, fully human antibodies, H6 and 7H, against human VCAM-1 (hVCAM-1) were developed and tested to validate the hypothesis that blocking VCAM-1 ameliorates atherosclerosis in apolipoprotein E-deficient (ApoE(-/-)) mice. METHODS AND RESULTS: Treatment with H6 or 7H effectively inhibited VCAM-1 adhesion to inflammatory cells, and reduced RhoA activation and the production of reactive oxygen species in human umbilical cord vascular endothelial cells. As 7H showed binding affinity to both murine VCAM-1 (mVCAM-1) and hVCAM-1, the therapeutic effects of 7H in ApoE(-/-) mice were tested. After confirming specific in vivo binding activity of 7H to mVCAM-1, we showed that administering 7H resulted in significantly ameliorated plaque formation compared to administering a control antibody in ApoE(-/-) mice fed a Western diet for 12 weeks. Also, 7H treatment significantly reduced infiltration of CD45(+) cells into plaques and reduced inflammation and improved plaque stability. CONCLUSION: These results indicate that the anti-VCAM-1 antibody attenuates atherosclerosis in ApoE(-/-) mice, improves plaque inflammation and stability as well as inhibiting the adhesion of inflammatory cell, and suggest that blocking VCAM-1 with a monoclonal antibody may be an effective means of anti-atherosclerotic therapy.

Ginkgo biloba extract (GbE) enhances the anti-atherogenic effect of cilostazol by inhibiting ROS generation.

In this study, the synergistic effect of 6-[4-(1-cyclohexyl- 1H-tetrazol-5-yl) butoxy]-3,4-dihydro-2(1H )-quinolinone (cilostazol) and Ginkgo biloba extract (GbE) was examined in apolipoprotein E (ApoE) null mice. Co-treatment with GbE and cilostazol synergistically decreased reactive oxygen species (ROS) production in ApoE null mice fed a high-fat diet. Co-treatment resulted in a significantly decreased atherosclerotic lesion area compared to untreated ApoE mice. The inflammatory cytokines and adhesion molecules such as monocyte chemoattractant-1 (MCP-1), soluble vascular cell adhesion molecule-1 (sVCAM-1), and VCAM-1 which can initiate atherosclerosis were significantly reduced by the co-treatment of cilostazol with GbE. Further, the infiltration of macrophages into the intima was decreased by co-treatment. These results suggest that co-treatment of GbE with cilostazol has a more potent anti-atherosclerotic effect than treatment with cilostazol alone in hyperlipidemic ApoE null mice and could be a valuable therapeutic strategy for the treatment of atherosclerosis.