Toward a structural identification of metastable molecular conformations.

Interpreting high-dimensional data from molecular dynamics simulations is a persistent challenge. In this paper, we show that for a small peptide, deca-alanine, metastable states can be identified through a neural net based on structural information alone. While processing molecular dynamics data, dimensionality reduction is a necessary step that projects high-dimensional data onto a low-dimensional representation that, ideally, captures the conformational changes in the underlying data. Conventional methods make use of the temporal information contained in trajectories generated through integrating the equations of motion, which forgoes more efficient sampling schemes. We demonstrate that EncoderMap, an autoencoder architecture with an additional distance metric, can find a suitable low-dimensional representation to identify long-lived molecular conformations using exclusively structural information. For deca-alanine, which exhibits several helix-forming pathways, we show that this approach allows us to combine simulations with different biasing forces and yields representations comparable in quality to other established methods. Our results contribute to computational strategies for the rapid automatic exploration of the configuration space of peptides and proteins.

Reagent prediction with a molecular transformer improves reaction data quality.

Automated synthesis planning is key for efficient generative chemistry. Since reactions of given reactants may yield different products depending on conditions such as the chemical context imposed by specific reagents, computer-aided synthesis planning should benefit from recommendations of reaction conditions. Traditional synthesis planning software, however, typically proposes reactions without specifying such conditions, relying on human organic chemists who know the conditions to carry out suggested reactions. In particular, reagent prediction for arbitrary reactions, a crucial aspect of condition recommendation, has been largely overlooked in cheminformatics until recently. Here we employ the Molecular Transformer, a state-of-the-art model for reaction prediction and single-step retrosynthesis, to tackle this problem. We train the model on the US patents dataset (USPTO) and test it on Reaxys to demonstrate its out-of-distribution generalization capabilities. Our reagent prediction model also improves the quality of product prediction: the Molecular Transformer is able to substitute the reagents in the noisy USPTO data with reagents that enable product prediction models to outperform those trained on plain USPTO. This makes it possible to improve upon the state-of-the-art in reaction product prediction on the USPTO MIT benchmark.

Future Speech Interfaces with Sensors and Machine Intelligence.

Speech is the most spontaneous and natural means of communication. Speech is also becoming the preferred modality for interacting with mobile or fixed electronic devices. However, speech interfaces have drawbacks, including a lack of user privacy; non-inclusivity for certain users; poor robustness in noisy conditions; and the difficulty of creating complex man-machine interfaces. To help address these problems, the Special Issue "Future Speech Interfaces with Sensors and Machine Intelligence" assembles eleven contributions covering multimodal and silent speech interfaces; lip reading applications; novel sensors for speech interfaces; and enhanced speech inclusivity tools for future speech interfaces. Short summaries of the articles are presented, followed by an overall evaluation. The success of this Special Issue has led to its being re-issued as "Future Speech Interfaces with Sensors and Machine Intelligence-II" with a deadline in March of 2023.

Analysis of Neural Network Based Proportional Myoelectric Hand Prosthesis Control.

OBJECTIVE: We show that state-of-the-art deep neural networks achieve superior results in regression-based multi-class proportional myoelectric hand prosthesis control than two common baseline approaches, and we analyze the neural network mapping to explain why this is the case. METHODS: Feedforward neural networks and baseline systems are trained on an offline corpus of 11 able-bodied subjects and 4 prosthesis wearers, using the R2 score as metric. Analysis is performed using diverse qualitative and quantitative approaches, followed by a rigorous evaluation. RESULTS: Our best neural networks have at least three hidden layers with at least 128 neurons per layer; smaller architectures, as used by many prior studies, perform substantially worse. The key to good performance is to both optimally regress the target movement, and to suppress spurious movements. Due to the properties of the underlying data, this is impossible to achieve with linear methods, but can be attained with high exactness using sufficiently large neural networks. CONCLUSION: Neural networks perform significantly better than common linear approaches in the given task, in particular when sufficiently large architectures are used. This can be explained by salient properties of the underlying data, and by theoretical and experimental analysis of the neural network mapping. SIGNIFICANCE: To the best of our knowledge, this work is the first one in the field which not only reports that large and deep neural networks are superior to existing architectures, but also explains this result.

User training for machine learning controlled upper limb prostheses: a serious game approach.

BACKGROUND: Upper limb prosthetics with multiple degrees of freedom (DoFs) are still mostly operated through the clinical standard Direct Control scheme. Machine learning control, on the other hand, allows controlling multiple DoFs although it requires separable and consistent electromyogram (EMG) patterns. Whereas user training can improve EMG pattern quality, conventional training methods might limit user potential. Training with serious games might lead to higher quality EMG patterns and better functional outcomes. In this explorative study we compare outcomes of serious game training with conventional training, and machine learning control with the users' own one DoF prosthesis. METHODS: Participants with upper limb absence participated in 7 training sessions where they learned to control a 3 DoF prosthesis with two grips which was fitted. Participants received either game training or conventional training. Conventional training was based on coaching, as described in the literature. Game-based training was conducted using two games that trained EMG pattern separability and functional use. Both groups also trained functional use with the prosthesis donned. The prosthesis system was controlled using a neural network regressor. Outcome measures were EMG metrics, number of DoFs used, the spherical subset of the Southampton Hand Assessment Procedure and the Clothespin Relocation Test. RESULTS: Eight participants were recruited and four completed the study. Training did not lead to consistent improvements in EMG pattern quality or functional use, but some participants improved in some metrics. No differences were observed between the groups. Participants achieved consistently better results using their own prosthesis than the machine-learning controlled prosthesis used in this study. CONCLUSION: Our explorative study showed in a small group of participants that serious game training seems to achieve similar results as conventional training. No consistent improvements were found in either group in terms of EMG metrics or functional use, which might be due to insufficient training. This study highlights the need for more research in user training for machine learning controlled prosthetics. In addition, this study contributes with more data comparing machine learning controlled prosthetics with Direct Controlled prosthetics.

Temporal dependency in automatic sleep scoring via deep learning based architectures: An empirical study.

The present study evaluates how effectively a deep learning based sleep scoring system does encode the temporal dependency from raw polysomnography signals. An exhaustive range of neural networks, including state of the art architecture, have been used in the evaluation. The architectures have been assessed using a single-channel EEG Fpz-Cz from the open source Sleep-EDF expanded database. The best performing model reached an overall accuracy of 85.2% and a Cohen's kappa of 0.8, with an F1-score of stage N1 equal to 50.2%. We have introduced a new metric, δnorm, to better evaluate temporal dependencies. A simple feed forward architecture not only achieves comparable performance to most up-to-date complex architectures, but also does better encode the continuous temporal characteristics of sleep.Clinical relevance - A better understanding of the capability of the network in encoding sleep temporal patterns could lead to improve the automatic sleep scoring.

Deep Learning Predicts Molecular Subtype of Muscle-invasive Bladder Cancer from Conventional Histopathological Slides.

BACKGROUND: Muscle-invasive bladder cancer (MIBC) is the second most common genitourinary malignancy, and is associated with high morbidity and mortality. Recently, molecular subtypes of MIBC have been identified, which have important clinical implications. OBJECTIVE: In the current study, we tried to predict the molecular subtype of MIBC samples from conventional histomorphology alone using deep learning. DESIGN, SETTING, AND PARTICIPANTS: Two cohorts of patients with MIBC were used: (1) The Cancer Genome Atlas Urothelial Bladder Carcinoma dataset including 407 patients and (2) our own cohort including 16 patients with treatment-naïve, primary resected MIBC. This resulted in a total of 423 digital whole slide images of tumor tissue to train, validate, and test the deep learning algorithm to predict the molecular subtype. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Various accuracy measurements including the area under the receiver operating characteristic curves were used to evaluate the deep learning model. A sliding window approach to visualize classification was used. Class activation maps were used to identify image features that are most relevant to call a specific class. RESULTS AND LIMITATIONS: The deep learning model showed great performance in the prediction of the molecular subtype of MIBC patients from hematoxylin and eosin (HE) slides alone-similar to or better than pathology experts. Using different visualization techniques, we identified new histopathological features that were most relevant to our model. CONCLUSIONS: Deep learning can be used to predict important molecular features in MIBC patients from HE slides alone, potentially improving the clinical management of this disease significantly. PATIENT SUMMARY: In patients with bladder cancer, a computer program found changes in the appearance of tumor tissue under the microscope and used these to predict genetic alterations. This could potentially benefit patients.

Electroclinic effect in a chiral carbosilane-terminated 5-phenylpyrimidine liquid crystal with 'de Vries-like' properties.

The chiral carbosilane-terminated liquid crystal 2-[(2S,3S)-2,3-difluorohexyloxy]-5-[4-(12,12,14,14,16,16-hexamethyl-12,14,16-trisilaheptadecyloxy)phenyl]pyrimidine () undergoes a smectic A*-smectic C* phase transition with a maximum layer contraction of only 0.2%. It exhibits an electroclinic effect (ECE) comparable to that reported for the 'de Vries-like' liquid crystal and shows no appreciable optical stripe defects due to horizontal chevron formation.

Tackling speaking mode varieties in EMG-based speech recognition.

An electromyographic (EMG) silent speech recognizer is a system that recognizes speech by capturing the electric potentials of the human articulatory muscles, thus enabling the user to communicate silently. After having established a baseline EMG-based continuous speech recognizer, in this paper, we investigate speaking mode variations, i.e., discrepancies between audible and silent speech that deteriorate recognition accuracy. We introduce multimode systems that allow seamless switching between audible and silent speech, investigate different measures which quantify speaking mode differences, and present the spectral mapping algorithm, which improves the word error rate (WER) on silent speech by up to 14.3% relative. Our best average silent speech WER is 34.7%, and our best WER on audibly spoken speech is 16.8%.

Pattern learning with deep neural networks in EMG-based speech recognition.

We report on classification of phones and phonetic features from facial electromyographic (EMG) data, within the context of our EMG-based Silent Speech interface. In this paper we show that a Deep Neural Network can be used to perform this classification task, yielding a significant improvement over conventional Gaussian Mixture models. Our central contribution is the visualization of patterns which are learned by the neural network. With increasing network depth, these patterns represent more and more intricate electromyographic activity.

Artifact removal algorithm for an EMG-based Silent Speech Interface.

An electromygraphic (EMG) Silent Speech Interface is a system which recognizes speech by capturing the electric potentials of the human articulatory muscles, thus enabling the user to communicate silently. This study deals with improving the EMG signal quality by removing artifacts: The EMG signals are captured by electrode arrays with multiple measuring points. On the resulting high-dimensional signal, Independent Component Analysis is performed, and artifact components are automatically detected and removed. This method reduces the Word Error Rate of the silent speech recognizer by 9.9% relative on a development corpus, and by 13.9% relative on an evaluation corpus.

Meshless Modeling of Deformable Shapes and their Motion.

We present a new framework for interactive shape deformation modeling and key frame interpolation based on a meshless finite element formulation. Starting from a coarse nodal sampling of an object's volume, we formulate rigidity and volume preservation constraints that are enforced to yield realistic shape deformations at interactive frame rates. Additionally, by specifying key frame poses of the deforming shape and optimizing the nodal displacements while targeting smooth interpolated motion, our algorithm extends to a motion planning framework for deformable objects. This allows reconstructing smooth and plausible deformable shape trajectories in the presence of possibly moving obstacles. The presented results illustrate that our framework can handle complex shapes at interactive rates and hence is a valuable tool for animators to realistically and efficiently model and interpolate deforming 3D shapes.

Ferroelectric liquid crystals induced by dopants with axially chiral 2,2'-spirobiindan-1,1'-dione cores.

The axially chiral dopants (R)-5,5'-, 5,6'-, and 6,6'-diheptyloxy-2,2'-spirobiindan-1,1'-dione ((R)-2, -3, and -4) were synthesized in optically pure form, and their absolute configurations were assigned by the exciton chirality method using circular dichroism spectroscopy. These new compounds were doped in four achiral liquid crystal hosts to give chiral smectic C* (SmC*) phases with spontaneous polarizations (Ps) that vary with the core structure of the host. The spontaneous polarization induced by the 5,5'-dialkoxy derivative (R)-2 is uniformly positive, whereas that induced by the 6,6'-dialkoxy derivative (R)-4 is uniformly negative and shows a different trend in host dependence. Polarization power (delta(p)) values range from +21 nC/cm2 for (R)-2 in 2',3'-difluoro-4-heptyl-4' '-nonyl-p-terphenyl (DFT) to -1037 nC/cm2 for (R)-4 in 4-(4'-heptyl[1,1'-biphen]-4-yl)-1-hexylcyclohexanecarbonitrile (NCB76). The unsymmetrical dopant (R)-3 behaves like a hybrid of the two symmetrical isomers, with lower absolute values of delta(p), on average, and varying signs of Ps. 2H NMR spectra of the doped mixtures using racemic mixtures of 2-4 with -OCD2C6H13 side-chains, in combination with phase diagrams, show that relatively minor changes in the dopant structure, that is, moving the alkoxy side-chains from the 5,5' to the 6,6' positions of the spirobiindandione core, have profound effects on dopant-host compatibility, and on the propensity of the dopant to exert chiral perturbations in the host environment. The variations in sign and magnitude of delta(p) as a function of alkoxy group positions are rationalized based on an analysis of zigzag conformations that conform to the binding site of the SmC host according to the Boulder model.

Photoinduced polarization inversion in a ferroelectric liquid crystal using an ambidextrous chiral thioindigo dopant.

In this Communication, we report the first example of photoswitching of a ferroelectric SmC* liquid crystal based on a photoinduced sign inversion of the spontaneous polarization (PS) induced by a single chiral dopant. This is achieved without concomitant destabilization of the SmC* phase using the "ambidextrous" dopant 6-((R,R)-2,3-difluorooct-1-yloxy)-5'-nitro-6'-((R)-2-octyloxy)thioindigo. The (R)-2-octyloxy side chain is sterically coupled to the thioindigo core via the nitro substituent and induces a positive polarization, whereas the (R,R)-2,3-difluorooct-1-yloxy side chain is decoupled from the core and induces a negative polarization. With this new design, the increase in transverse dipole moment of the thioindigo core upon trans-cis photoisomerization raises the polarization power of the coupled 2-octyloxy/thioindigo unit above that of the 2,3-difluorooctyloxy unit and inverts the net sign of PS.

Detection of chiral perturbations in ferroelectric liquid crystals induced by an atropisomeric biphenyl dopant.

The atropisomeric dopant 2,2',6,6'-tetramethyl-3,3'-dinitro-4,4'-bis[(4-nonyloxybenzoyl)oxy]biphenyl (1) induces a ferroelectric SmC phase when doped into the SmC liquid crystal hosts 2-(4-butyloxyphenyl)-5-octyloxypyrimidine (PhP1) and (+/-)-4-[(4-methylhexyl)oxy]phenyl 4-decyloxybenzoate (PhB). The propensity of dopant 1 to induce a spontaneous polarization (polarization power) is much higher in PhP1 than in PhB (1555 nC/cm(2) vs <35 nC/cm(2)), which is attributed to a greater propensity of 1 to undergo chirality transfer via core-core interactions with PhP1. In previous work, we postulated that a chiral perturbation exerted by 1 in PhP1 amplifies the polarization power of the dopant by causing a chiral distortion of the mean field potential (binding site) constraining the dopant in the SmC host, as described by the Chirality Transfer Feedback (CTF) model. To test the validity of the CTF model, and to provide a more direct assessment of the chiral perturbation exerted by dopant 1 on surrounding host molecules, we measured the effect of 1 on the polarization power of other chiral dopants acting as probes. In one series of experiments, (S,S)-5-(2,3-difluorooctyl)-2-(4-octylphenyl)pyridine (MDW950) and (S)-4-(1-methylheptyloxy)phenyl 4-decyloxybenzoate (4), which mimic the structures of PhP1 and PhB, were used as probes. In another series of experiments, the atropisomeric dopant 2,2',3,3',6,6'-hexamethyl-4,4'-bis[(4-nonyloxybenzoyl)oxy]biphenyl (2) was used as probe in PhP1. The results of the probe experiments suggest that dopant 1 exerts a much stronger chiral perturbation in PhP1 than in PhB. More significantly, the results of experiments using 2 as probe show that the chiral perturbation exerted by 1 can amplify the polarization power of another atropisomeric dopant, thus providing the first experimental evidence of the CTF effect.

Bistable ferroelectric liquid crystal photoswitch triggered by a dithienylethene dopant.

The spontaneous polarization (PS) of a ferroelectric liquid crystal is modulated reversibly by photocyclization of the dopant 1,2-bis[5'-(4' '-heptyloxyphenyl)-2'-methylthien-3'-yl]perfluorocyclopentene. The magnitude of PS photomodulation increases with dopant concentration up to 3 mol %, and the resulting photoswitch is fatigue resistant and bistable. To the best of our knowledge, this is the first example of a bistable ferroelectric liquid crystal photoswitch to be reported in the literature.