Weijie Chen

Objective: Inter-session and inter-subject variability in electroencephalography (EEG) signals, resulting from individual differences and environmental factors, poses a major challenge for neural decoding in brain-computer interface (BCI) applications. Methods: To address this issue, we propose RUNet, a zero-calibration motor imagery EEG decoding framework based on Riemannian manifold learning and unsupervised representation learning. RUNet incorporates a multi-scale spatiotemporal convolutional module that jointly captures local global spatial and multi-resolution temporal dynamics features. To enhance the robustness of EEG features against non stationarity, a polysynergistic covariance optimization module is employed, which strengthens the covariance matrix representation through multiple regularizations and adaptive fusion. In addition, RUNet integrates the Riemannian Affine Log Mapping layer, based on Affine-Invariant Transformation and Log-Euclidean Mapping, in an end-to-end manner to mitigate cross-domain covariance drift and enhance domain-invariant feature learning. A transfer learning framework is further integrated into RUNet: during pre-training, an unsupervised contrastive loss is applied to resting-state EEG data to learn domain-invariant spatiotemporal features; during retraining, task-specific data are used to enhance discriminability and feature disentanglement. Conclusion: Experimental results on the BCI Competition IV 2a, 2b datasets and a self-collected laboratory dataset show that RUNet achieves average cross-session accuracies of 87.19%, 88.03% and 85.45%, and cross-subject accuracies of 68.09%, 78.29% and 87.25%, respectively. On the PhysioNet dataset, a cross-subject accuracy of 78.14% is achieved. These results demonstrate the effectiveness of RUNet's unified pipeline and its robust cross-domain generalization.

Supplementary Figure S1 illustrates the chemical structures of SCR-6920, GSK3235025, and LLY-283 (panel A) and overlays their binding poses in the PRMT5 active site, highlighting key interactions with SAM and the protein (panels B–E)

Supplementary Figure S3 shows the effects of SCR-6920 on tumor growth and body weight in mouse xenograft models (Z138, 5637, and MDA-MB-468), demonstrating dose-dependent tumor inhibition without significant body weight loss compared to vehicle and GSK3326595 controls.

Supplementary Figure S4 shows that high PRMT5 expression correlates with poor overall survival in cancer patients, and demonstrates that SCR-6920 enhances protein ubiquitination under hypoxia while synergizing with chemotherapies (docetaxel, paclitaxel + Avastin) to suppress tumor growth without affecting body weight in xenograft models.

Protein arginine methyltransferase 5 (PRMT5) is a critical therapeutic target in cancer. In this study, we report SCR-6920, a potent and selective oral PRMT5 inhibitor characterized by a substrate-competitive binding mode to PRMT5:methylosome protein 50 with high affinity. SCR-6920 shows broad antitumor activity in vitro and in vivo, along with favorable pharmacokinetics and tumor-biased distribution. Notably, RNA sequencing analysis links vascular endothelial growth factor (VEGF) to PRMT5, and further studies show that SCR-6920 downregulates hypoxia-inducible factor-1α (HIF-1α) through the ubiquitin–proteasome degradation pathway, thereby reducing downstream VEGF expression and secretion. Thus, PRMT5 inhibition enhances hypoxia-induced pathway disruption, and a further in vivo efficacy study demonstrates a synergistic effect for SCR-6920 combined with bevacizumab against ovarian cancer. Additionally, SCR-6920 improves the efficacy of clinical first-line ovarian cancer therapies (paclitaxel, docetaxel, doxorubicin, and olaparib). This study identifies a promising PRMT5 inhibitor with synergistic activity with bevacizumab and reveals its regulation of HIF-1α signaling, highlighting the therapeutic potential of PRMT5 inhibition in combination regimens for ovarian cancer.

Supplementary Figure S2 shows per-residue energy decomposition for SCR-6920 and GSK3235025 bound to PRMT5, comparing van der Waals, non-polar solvation, electrostatic, and total interaction energies across the binding pocket residues.

Supplementary Table S1 lists various cancer cell lines used in the study, including lung, ovarian, breast, head and neck, bladder, and lymphoma cell lines. For each cell line, it records information such as cancer type, cell line name, vendor, RRID, sex, acquisition year, STR authentication date, mycoplasma test date, culture medium, and IC50 values.

Supplementary Table S2 summarizes the off-target profiling of SCR-6920 across 92 assays, showing no significant activity (RC50 > 10 µM) against GPCRs, ion channels, kinases, nuclear hormone receptors, enzymes, and transporters.

Supplementary Table S6 summarizes the pharmacokinetic parameters of SCR-6920 in whole blood of Z138 tumor-bearing mice across three dose levels (2.5, 7.5, and 15 mg/kg), showing dose-proportional increases in Cmax and AUClast, consistent tmax of 0.5 hours, and decreasing half-life with increasing dose.

Supplementary Table S7 summarizes the pharmacokinetic parameters of SCR-6920 in rats and dogs after single intravenous and oral doses, showing moderate oral bioavailability (≈40 %) and longer half-life in dogs compared to rats.