SORBET: Automated cell-neighborhood analysis of spatial transcriptomics or proteomics for interpretable sample classification via GNN.

Spatially resolved transcriptomics or proteomics data have the potential to contribute fundamental insights into the mechanisms underlying physiologic and pathological processes. However, analysis of these data capable of relating spatial information, multiplexed markers, and their observed phenotypes remains technically challenging. To analyze these relationships, we developed SORBET, a deep learning framework that leverages recent advances in graph neural networks (GNN). We apply SORBET to predict tissue phenotypes, such as response to immunotherapy, across different disease processes and different technologies including both spatial proteomics and transcriptomics methods. Our results show that SORBET accurately learns biologically meaningful relationships across distinct tissue structures and data acquisition methods. Furthermore, we demonstrate that SORBET facilitates understanding of the spatially-resolved biological mechanisms underlying the inferred phenotypes. In sum, our method facilitates mapping between the rich spatial and marker information acquired from spatial 'omics technologies to emergent biological phenotypes. Moreover, we provide novel techniques for identifying the biological processes that comprise the predicted phenotypes.

Biophysical prediction of protein-peptide interactions and signaling networks using machine learning.

In mammalian cells, much of signal transduction is mediated by weak protein-protein interactions between globular peptide-binding domains (PBDs) and unstructured peptidic motifs in partner proteins. The number and diversity of these PBDs (over 1,800 are known), their low binding affinities and the sensitivity of binding properties to minor sequence variation represent a substantial challenge to experimental and computational analysis of PBD specificity and the networks PBDs create. Here, we introduce a bespoke machine-learning approach, hierarchical statistical mechanical modeling (HSM), capable of accurately predicting the affinities of PBD-peptide interactions across multiple protein families. By synthesizing biophysical priors within a modern machine-learning framework, HSM outperforms existing computational methods and high-throughput experimental assays. HSM models are interpretable in familiar biophysical terms at three spatial scales: the energetics of protein-peptide binding, the multidentate organization of protein-protein interactions and the global architecture of signaling networks.

Unexpected right hemisphere language representation identified by the intracarotid amobarbital procedure in right-handed epilepsy surgery candidates.

The intracarotid amobarbital procedure (IAP) is used for language lateralization in the presurgical evaluation for intractable epilepsy. Some epilepsy surgical centers forgo IAP for right temporal lobectomies in right-handed patients with no personal/family history of left-handedness, implying that right hemisphere language representation does not occur in these patients. To test this hypothesis, a retrospective analysis was performed on 156 consecutive epilepsy surgery candidates who underwent IAP. Of the 156 candidates, 122 were right-handed, and 55 of the 122 demonstrated right hemisphere seizure focus. Right hemisphere language representation was found in 22 of 55 patients, with two demonstrating significant right hemisphere language despite a right hemisphere seizure focus and no family history of left-handedness. Only 1 of 156 patients undergoing IAP experienced permanent neurological complications. Although relatively uncommon, right hemisphere language representation may occur more frequently than complications from cerebral angiography and, therefore, presurgical IAP is recommended for all epilepsy surgery candidates regardless of handedness to minimize the risk of severe language decline.

Neural basis of the Stroop interference task: response competition or selective attention?

Previous neuroimaging studies of the Stroop task have postulated that the anterior cingulate cortex (ACC) plays a critical role in resolution of the Stroop interference condition. However, activation of the ACC is not invariably seen and appears to depend on a variety of methodological factors, including the degree of response conflict and response expectancies. The present functional MRI study was designed to identify those brain areas critically involved in the interference condition. Healthy subjects underwent a blocked-trial design fMRI experiment while responding to 1 of 3 stimulus conditions: (1) incongruent color words, (2) congruent color words, and (3) color-neutral words. Subjects responded to the printed color of the word via a manual response. Compared to the congruent and neutral conditions, the incongruent condition produced significant activation within the left inferior precentral sulcus (IpreCS) located on the border between the inferior frontal gyrus, pars opercularis (BA 44) and the ventral premotor region (BA 6). Significant deactivations in the rostral component of the ACC and the posterior cingulate gyrus were also observed. Selective activation of the left IpreCS is compatible with findings from previous neuroimaging, lesion, electrophysiological, and behavioral studies and is presumably related to the mediation of competing articulatory demands during the interference condition.