The integration of gene expression data to predict systemic lupus erythematosus (SLE) disease activity is a significant challenge because of the high degree of heterogeneity among patients and study cohorts, especially those collected on different microarray platforms. Here we deployed machine learning approaches to integrate gene expression data from three SLE data sets and used it to classify patients as having active or inactive disease as characterized by standard clinical composite outcome measures. Both raw whole blood gene expression data and informative gene modules generated by Weighted Gene Co-expression Network Analysis from purified leukocyte populations were employed with various classification algorithms. Classifiers were evaluated by 10-fold cross-validation across three combined data sets or by training and testing in independent data sets, the latter of which amplified the effects of technical variation. A random forest classifier achieved a peak classification accuracy of 83 percent under 10-fold cross-validation, but its performance could be severely affected by technical variation among data sets. The use of gene modules rather than raw gene expression was more robust, achieving classification accuracies of approximately 70 percent regardless of how the training and testing sets were formed. Fine-tuning the algorithms and parameter sets may generate sufficient accuracy to be informative as a standalone estimate of disease activity.

The zinc-binding Bbox1 domain in protein MID1, a member of the TRIM family of proteins, facilitates the ubiquitination of the catalytic subunit of protein phosphatase 2A and alpha4, a protein regulator of PP2A. The natural mutation of residue A130 to a valine or threonine disrupts substrate recognition and catalysis. While NMR data revealed the A130T mutant Bbox1 domain failed to coordinate both structurally essential zinc ions and resulted in an unfolded structure, the unfolding mechanism is unknown. Principle component analysis revealed that residue A130 served as a hinge point between the structured β-strand-turn-β-strand (β-turn-β) and the lasso-like loop sub-structures that constitute loop1 of the ββα-RING fold that the Bbox1 domain adopts. Backbone RMSD data indicate significant flexibility and departure from the native structure within the first 5 ns of the molecular dynamics (MD) simulation for the A130V mutant (>6 Å) and after 30 ns for A130T mutant (>6 Å). Overall RMSF values were higher for the mutant structures and showed increased flexibility around residues 125 and 155, regions with zinc-coordinating residues. Simulated pKa values of the sulfhydryl group of C142 located near A130 suggested an increased in value to ~9.0, paralleling the increase in the apparent dielectric constants for the small cavity near residue A130. Protonation of the sulfhydryl group would disrupt zinc-coordination, directly contributing to unfolding of the Bbox1. Together, the increased motion of residues of loop 1, which contains four of the six zinc-binding cysteine residues, and the increased pKa of C142 could destabilize the structure of the zinc-coordinating residues and contribute to the unfolding.

Sociability is crucial for survival, whereas social avoidance is a feature of disorders such as Rett syndrome, which is caused by loss-of-function mutations in MECP2. To understand how a preference for social interactions is encoded, we used in vivo calcium imaging to compare medial prefrontal cortex (mPFC) activity in female wild-type and Mecp2-heterozygous mice during three-chamber tests. We found that mPFC pyramidal neurons in Mecp2-deficient mice are hypo-responsive to both social and nonsocial stimuli. Hypothesizing that this limited dynamic range restricts the circuit’s ability to disambiguate coactivity patterns for different stimuli, we suppressed the mPFC in wild-type mice and found that this eliminated both pattern decorrelation and social preference. Conversely, stimulating the mPFC in MeCP2-deficient mice restored social preference, but only if it was sufficient to restore pattern decorrelation. A loss of social preference could thus indicate impaired pattern decorrelation rather than true social avoidance.

RNA-protein complexes underlie numerous cellular processes, including basic translation and gene regulation. The high-resolution structure determination of the RNA-protein complexes is essential for elucidating their functions. Therefore, computational methods capable of identifying the native-like RNA-protein structures are needed. To address this challenge, we thus develop DRPScore, a deep-learning-based approach for identifying native-like RNA-protein structures. DRPScore is tested on representative sets of RNA-protein complexes with various degrees of binding-induced conformation change ranging from fully rigid docking (bound-bound) to fully flexible docking (unbound-unbound). Out of the top 20 predictions, DRPScore selects native-like structures with a success rate of 91.67% on the testing set of bound RNA-protein complexes and 56.14% on the unbound complexes. DRPScore consistently outperforms existing methods with a roughly 10.53–15.79% improvement, even for the most difficult unbound cases. Furthermore, DRPScore significantly improves the accuracy of the native interface interaction predictions. DRPScore should be broadly useful for modeling and designing RNA-protein complexes.

A central challenge in systems biology today is to understand the network of interactions among biomolecules and, especially, the organizing principles underlying such networks. Recent analysis of known networks has identified small motifs that occur ubiquitously, suggesting that larger networks might be constructed in the manner of electronic circuits by assembling groups of these smaller modules. Using a unique process-based approach to analyzing such networks, we show for two cell-cycle networks that each of these networks contains a giant backbone motif spanning all the network nodes that provides the main functional response. The backbone is in fact the smallest network capable of providing the desired functionality. Furthermore, the remaining edges in the network form smaller motifs whose role is to confer stability properties rather than provide function. The process-based approach used in the above analysis has additional benefits: It is scalable, analytic (resulting in a single analyzable expression that describes the behavior), and computationally efficient (all possible minimal networks for a biological process can be identified and enumerated).

Accurate computational determination of RNA-protein interactions remains challenging, particularly when encountering unknown RNAs and proteins. The limited number of RNAs and their flexibility constrained the effectiveness of the deep-learning models for RNA-protein interaction prediction. Here, we introduce ZHMolGraph, which integrates graph neural network and unsupervised large language models to predict RNA-protein interaction. We validate ZHMolGraph predictions on two benchmark datasets and outperform the current best methods. For the dataset of entirely unknown RNAs and proteins, ZHMolGraph shows an improvement in achieving high AUROC of 79.8% and AUPRC of 82.0%. This represents a substantial improvement of 7.1%–28.7% in AUROC and 4.6%–30.0% in AUPRC over other methods. We utilize ZHMolGraph to enhance the challenging SARS-CoV-2 RPI and unbound RNA-protein complex predictions. Such enhancements make ZHMolGraph a reliable option for genome-wide RNA-protein prediction. ZHMolGraph holds broad potential for modeling and designing RNA-protein complexes.