Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

Predicting falls and injuries in people with multiple sclerosis using machine learning algorithms.

In Multiple sclerosis and related disorders

Falls in people with Multiple Sclerosis (PwMS) is a serious issue. It can lead to a lot of problems including sustaining injuries, losing consciousness and hospitalization. Having a model that can predict the probability of these falls and the factors correlated with them and can help caregivers and family members to have a clearer understanding of the risks of falling and proactively minimizing them. We used historical data and machine learning algorithms to predict three outcomes: falling, sustaining injuries and injury types caused by falling in PwMS. The training dataset for this study includes 606 examples of monthly readings. The predictive attributes are the following: Expanded Disability Status Scale (EDSS), years passed since the diagnosis of MS, age of participants in the beginning of the experiment, participants' gender, type of MS and season (or month). Two types of algorithms, decision tree and gradient boosted trees (GBT) algorithm, were used to train six models to answer these three outcomes. After the models were trained their accuracy was evaluated using cross-validation. The models had a high accuracy with some exceeding 90%. We did not limit model evaluation to one-number assessments and studied the confusion matrices of the models as well. The GBT had a higher class recall and smaller number of underestimations, which make it a more reliable model. The methodology proposed in this study and its findings can help in developing better decision-support tools to assist PwMS.

Piryonesi S Madeh, Rostampour Sorour, Piryonesi S Abdurrahman


Fall prediction, Injury, Machine learning, Model evaluation, Multiple sclerosis

General General

The relevance to social interaction modulates bistable biological-motion perception.

In Cognition

Social interaction, the process through which individuals act and react toward each other, is arguably the building block of society. As the very first step for successful social interaction, we need to derive the orientation and immediate social relevance of other people: a person facing toward us is much more likely to initiate communications than a person who is back to us. Reversely, however, it remains elusive whether the relevance to social interaction modulates how we perceive the other's orientation. Here, we adopted the bistable point-light walker (PLW) which is ambiguous in its in-depth orientation. Participants were asked to report the orientation (facing the viewer or facing away from the viewer) of the PLWs. Three factors that are task-irrelevant but critically pertinent to social interaction, the distance, the speed, and the size of the PLW, were systematically manipulated. The nearer a person is, the more likely it initiates interactions with us. The larger a person is, the larger influence it may exert. The faster a person is, the shorter time is left for us to respond. Results revealed that participants tended to perceive the PLW as facing them more frequently than facing away when the PLW was nearer, faster, or larger. These same factors produced different patterns of effects on a non-biological rotating cylinder. These findings demonstrate that the relevance to social interaction modulates the visual perception of biological motion and highlight that bistable biological motion perception not only reflects competitions of low-level features but is also strongly linked to high-level social cognition.

Han Qiu, Wang Ying, Jiang Yi, Bao Min


Biological motion, Bistable perception, Facing bias, Social cognition, Social interaction

General General

ELASPIC2 (EL2): Combining contextualized language models and graph neural networks to predict effects of mutations.

In Journal of molecular biology ; h5-index 65.0

The ELASPIC web server allows users to evaluate the effect of mutations on protein folding and protein-protein interaction on a proteome-wide scale. It uses homology models of proteins and protein-protein interactions, which have been precalculated for several proteomes, and machine learning models, which integrate structural information with sequence conservation scores, in order to make its predictions. Since the original publication of the ELASPIC web server, several advances have motivated a revisiting of the problem of mutation effect prediction. First, progress in neural network architectures and self-supervised pre-trained has resulted in models which provide more informative embeddings of protein sequence and structure than those used by the original version of ELASPIC. Second, the amount of training data has increased several-fold, largely driven by advances in deep mutation scanning and other multiplexed assays of variant effect. Here, we describe two machine learning models which leverage the recent advances in order to achieve superior accuracy in predicting the effect of mutation on protein folding and protein-protein interaction. The models incorporate features generated using pre-trained transformer- and graph convolution-based neural networks, and are trained to optimize a ranking objective function, which permits the use of heterogeneous training data. The outputs from the new models have been incorporated into the ELASPIC web server, available at

Strokach Alexey, Yu Lu Tian, Kim Philip M


Variant effect prediction, affinity prediction, graph convolutional neural network, machine learning, stability prediction

General General

Brain lipidomics as a rising field in neurodegenerative contexts: Perspectives with Machine Learning approaches.

In Frontiers in neuroendocrinology ; h5-index 44.0

Lipids are essential for cellular functioning considering their role in membrane composition, signaling, and energy metabolism. The brain is the second most abundant organ in terms of lipid concentration and diversity only after adipose tissue. However, in the central system (CNS) lipid dysregulation has been linked to the etiology, progression, and severity of neurodegenerative diseases such as Alzheimeŕs, Parkinson, and Multiple Sclerosis. Advances in the human genome and subsequent sequencing technologies allowed us the study of lipidomics as a promising approach to diagnosis and treatment of neurodegeneration. Lipidomics advances rapidly increased the amount and quality of data allowing the integration with other omic types as well as implementing novel bioinformatic and quantitative tools such as machine learning (ML). Integration of lipidomics data with ML, as a powerful quantitative predictive approach, led to improvements in diagnostic biomarker prediction, clinical data integration, network, and systems approaches for neural behavior, novel etiology markers for inflammation, and neurodegeneration progression and even Mass Spectrometry image analysis. In this sense, by exploiting lipidomics data with ML is possible to improve the identification of new biomarkers or unveil new molecular mechanisms associated with lipid impairment across neurodegeneration. In this review, we present the lipidomic neurobiology state-of-the-art highlighting its potential applications to study neurodegenerative conditions. Also, we present theoretical background, applications, and advances in the integration of lipidomics with ML. This review opens the door to new approaches in this rising field.

Báez Castellanos Daniel, Martín-Jiménez Cynthia A, Rojas-Rodríguez Felipe, Barreto George, González Santos Janneth


Alzheimer’s Disease(5), Fatty acids(4), Lipidomics(2), Machine Learning(3), Multiple Sclerosis(7), Neurodegeneration(1), Parkinson Disease(6)

General General

Machine Learning Applied to Determine the Molecular Descriptors Responsible for the Viscosity Behavior of Concentrated Therapeutic Antibodies.

In Molecular pharmaceutics ; h5-index 60.0

Predicting the solution viscosity of monoclonal antibody (mAb) drug products remains as one of the main challenges in antibody drug design, manufacturing, and delivery. In this work, the concentration-dependent solution viscosity of 27 FDA-approved mAbs was measured at pH 6.0 in 10 mM histidine-HCl. Six mAbs exhibited high viscosity (>30 cP) in solutions at 150 mg/mL mAb concentration. Combining molecular modeling and machine learning feature selection, we found that the net charge in the mAbs and the amino acid composition in the Fv region are key features which govern the viscosity behavior. For mAbs whose behavior was not dominated by charge effects, we observed that high viscosity is correlated with more hydrophilic and fewer hydrophobic residues in the Fv region. A predictive model based on the net charges of mAbs and a high viscosity index is presented as a fast screening tool for classifying low- and high-viscosity mAbs.

Lai Pin-Kuang, Fernando Amendra, Cloutier Theresa K, Gokarn Yatin, Zhang Jifeng, Schwenger Walter, Chari Ravi, Calero-Rubio Cesar, Trout Bernhardt L


intermolecular interactions, machine learning, molecular modeling, therapeutic antibodies, viscosity

General General

A systems-level gene regulatory network model for Plasmodium falciparum.

In Nucleic acids research ; h5-index 217.0

Many of the gene regulatory processes of Plasmodium falciparum, the deadliest malaria parasite, remain poorly understood. To develop a comprehensive guide for exploring this organism's gene regulatory network, we generated a systems-level model of P. falciparum gene regulation using a well-validated, machine-learning approach for predicting interactions between transcription regulators and their targets. The resulting network accurately predicts expression levels of transcriptionally coherent gene regulatory programs in independent transcriptomic data sets from parasites collected by different research groups in diverse laboratory and field settings. Thus, our results indicate that our gene regulatory model has predictive power and utility as a hypothesis-generating tool for illuminating clinically relevant gene regulatory mechanisms within P. falciparum. Using the set of regulatory programs we identified, we also investigated correlates of artemisinin resistance based on gene expression coherence. We report that resistance is associated with incoherent expression across many regulatory programs, including those controlling genes associated with erythrocyte-host engagement. These results suggest that parasite populations with reduced artemisinin sensitivity are more transcriptionally heterogenous. This pattern is consistent with a model where the parasite utilizes bet-hedging strategies to diversify the population, rendering a subpopulation more able to navigate drug treatment.

Neal Maxwell L, Wei Ling, Peterson Eliza, Arrieta-Ortiz Mario L, Danziger Samuel A, Baliga Nitin S, Kaushansky Alexis, Aitchison John D