MicroRNAs (miRNAs) are key post-transcriptional regulators closely associated with human diseases. Identifying miRNA-drug associations (MDAs) is important for precision drug discovery, yet experimental validation remains costly and labor-intensive. To address this limitation, we propose MVR-MDA, a multi-modal deep learning framework that integrates heterogeneous biological information for efficient MDA prediction. MVR-MDA combines three complementary feature sources: pretrained 3D molecular representations from Uni-Mol to enhance generalization to novel drugs, intra-attribute features from MACCS fingerprints and miRNA sequences refined by BiGRU, and inter-topological features captured through SDNE from the miRNA-drug interaction graph. These representations are fused to learn both intrinsic molecular properties and global relational patterns. We evaluate MVR-MDA on ncDR and RNAInter through 5-fold cross-validation, achieving superior prediction performance compared to state-of-the-art methods. A case study on 5-Fluorouracil and hsa-miR-146a further validates the biological relevance of our predictions. Overall, MVR-MDA provides an effective computational tool for discovering potential MDAs, supporting novel therapeutic target identification and accelerating drug repositioning.

Teleoperation is a crucial tool for collecting human demonstrations, but controlling robots with bimanual dexterous hands remains a challenge. Existing teleoperation systems struggle to handle the complexity of coordinating two hands for intricate manipulations. We introduce Bunny-VisionPro, a real-time bimanual dexterous teleoperation system that leverages a VR headset. Unlike previous vision-based teleoperation systems, we design novel low-cost devices to provide haptic feedback to the operator, enhancing immersion. Our system prioritizes safety by incorporating collision and singularity avoidance while maintaining real-time performance through innovative designs. Bunny-VisionPro outperforms prior systems on a standard task suite, achieving higher success rates and reduced task completion times. Moreover, the high-quality teleoperation demonstrations improve downstream imitation learning performance, leading to better generalizability. Notably, Bunny-VisionPro enables imitation learning with challenging multi-stage, long-horizon dexterous manipulation tasks, which have rarely been addressed in previous work. Our system’s ability to handle bimanual manipulations while prioritizing safety and real-time performance makes it a powerful tool for advancing dexterous manipulation and imitation learning. Our web page is available at https://dingry.github.io/projects/bunny_visionpro.

Backscattering from rough soil surface has important applications to the remote sensing of soil moisture and snow water equivalent (SWE), To extend simulations to Ku-band, we have performed full wave simulations up tokh=15. Results are simulated for various profiles and show that the scattering at C-, X-, and Ku-bands is influenced by both scales of centimeters roughness and millimeter roughness. Comparisons are made between constant ratios rough surface and constant correlation length rough surfaces. The results are illustrated for both VV and HH polarizations. Simulation results are in good agreement with X band measurement data as a function of incidence angle. An illustration is used to show how the results can be used for theoretical models of rough surface scattering in remote sensing of snow water equivalent.

A comparison of three different electromagnetic scattering models for land surface delay-Doppler maps (DDMs) obtained from global navigation satellite system reflectometry (GNSS-R) along a Cyclone Global Navigation Satellite System (CYGNSS) track in the San Luis Valley, Colorado, USA, is presented. The three models are the analytical Kirchhoff solutions (AKS), the Soil And VEgetation Reflection Simulator (SAVERS), and the improved geometrical optics with topography (IGOT). Common inputs to the three models were defined by using field samples of soil moisture and texture, soil surface roughness measurements, and a digital elevation model (DEM). The resulting peak reflectivity profiles of the models and the CYGNSS data all had a range of 10 dB along the selected track, mainly due to the influence of topography. The reflectivities obtained from all three models agreed with one another to within 2.4 dB along the full length of the track. The models also showed general agreement with the corresponding CYGNSS data, although the modeled profiles were higher than CYGNSS Science Data Record Version 3.1 by an average of 5 dB and also smoother. Additional characterization of fine-scale surface roughness is identified as an area for future work to improve model fidelity. An intercomparison of DDM structure for three selected acquisitions is also provided.

A comparison of three different electromagnetic scattering models for delay-Doppler maps (DDMs) of global navigation satellite system reflectometry (GNSS-R) from land is performed along a Cyclone Global Navigation Satellite System (CYGNSS) track over a validation site in the San Luis Valley, Colorado, USA. The peak reflectivity profiles of all three models and of the corresponding CYGNSS data are found to be in general agreement and are strongly influenced by topography. An intercomparison of DDM structure for one acquisition is also included. Efforts to refine the model results using a high resolution lidar survey are ongoing.

In this paper we present processing and analysis results of an airborne Ku-band InSAR, constructed at the University of Massachusetts, and flown on a Cessna 208 Caravan over the Trail Valley Creek region in Canada's Northwest Territories during the 2018–19 snow season. In this paper, we describe the Ku-band InSAR, provide some intermediate results and discuss on how these data can be used for furthering the science in the remote sensing of snow.

The capability of Sentinel 1 C band radar observations for mapping snow depth or snow water equivalent (SWE) has been demonstrated recently. However, theoretical models of C band radar signatures for snow retrieval are still lacking. In this paper, we study the volume scattering of snowpack and the surface scattering from the snow/soil interface. The snowpack is computer generated including dense ice aggregates. The volume scattering is calculated by the dense media radiative transfer (DMRT) model and the surface scattering is computed with the Oh model. With well characterized surface scattering, surface scattering contributions can be subtracted from radar observations to enhance sensitivity of volume scattering to SWE. The study will help improve C band SWE retrieval and provide the theoretical basis for retrieval algorithms.

In this paper, we study the scattering of land surfaces for the Global Navigation Satellite System Reflectometry (GNSS-R) land applications. Topography slopes are introduced to improve the physical patch model. The entire area within the footprint is divided into patches. Each patch is on a slope with elevation. Simulation results have shown that for small rms height, the received power to transmitted power ratio, Pr/Pt is smaller than the coherent model and the patch model with only elevation effects but greater than the incoherent model. The delay-doppler maps are also simulated based on the patch model.

Previous experiments showed that the frequency dependence of snow volume scattering from 18 to 90GHz is power of 2.8, which is much weaker than the power of 4 of Rayleigh scattering. Recently monitoring snow using existing satellites (such as Sentinel-1, COSMO-SkyMed and QuickScat) have been widely studied. The radar signatures from C to Ku band are important for active remote sensing of snow. In this paper, we study the snow aggregation effects and its frequency dependence of volume scattering from C to Ku band (4-18GHz) with the bicontinuous model. The bicontinuous media model is applied to model the snow microstructure with aggregates. The integral equation of snow scattering volume is derived based on the Born approximation which then leads to the scattering coefficients. Scattering coefficients of snow are computed from C to Ku band giving a frequency dependence of ~2.6 power, weaker than that of Rayleigh scattering because of the aggregation effects. In addition, results are also in good agreement with the full wave numerical solutions from X to Ku band.

Arctic and Antarctic sea ice covers are in a sharp contrast in terms of characteristics, distributions, and processes with a drastic decrease in the Arctic versus the opposite increase in the Antarctic in a changing climate. In quantifying polar sea ice differences to address the contrasted sea ice behaviors, two key parameters are sea ice thickness and melt pond fraction, which remain challenging to measure extensively in time and in space with a sustainable approach. Here, we present a new paradigm for such measurements using bistatic radar reflectometry, thanks to developments of low-cost receivers to acquire reflected signals from numerous existing transmitter systems operated at multiple frequencies to be replenished and sustained indefinitely into the future. For sea ice thickness measurement to determine ice volume, reflected signals likely come from the bottom ice-water interface avoiding large errors inherent in current altimetry techniques due to uncertainty in free-board height and snow cover. Regarding melt pond faction on sea ice to estimate albedo and insolation, the bistatic reflection can be dominated by melt pond water with permittivity that is one order of magnitude larger compared to that of snow or ice. These are examined by a combination of numerical Kirchhoff (KA) simulator and Numerical Maxwell Model of 3D simulations (NMM3D) to preserve phase and amplitude information and thereby account for both coherent and incoherent effects. Physical insights from the rigorous theory for bistatic radar reflectometry will be valuable to develop future satellite missions to resolve cryospheric science issues concerning the polar sea ice differences.

The Global Navigation Satellite System Reflectometry (GNSS-R) for land applications have attracted considerable attention. Prior models include the coherent model and incoherent models. There are big differences between these two models. In this paper, we propose a patch model with NMM3D (Numerical Maxwell model of 3D simulations). The models take effects of multiple elevations of land surfaces into account. Results are compared with 1) coherent model, 2) incoherent model, and (3) numerical Kirchhoff simulator (KA simulator). Results of the patch model are consistent with those of KA simulator and have many dB difference with coherent and incoherent model.

We apply the analytical fully coherent model to bicontinuous media for applications in microwave remote sensing of snow cover. In our model, snow is represented by bicontinuous media which is generated by a Gaussian random process. The statistical moments and correlation functions of bicontinuous media can be calculated. With the correlation functions, we apply the strong permittivity fluctuation (SPF) theory to derive the analytical solutions for calculation of scattering properties and effective permittivity of bicontinuous meida. The SPF theory is under the bilocal approximation. Effective permittivity and extinction coefficient from the analytical solutions are compared with those of the numerical solution of Maxwell's equation in 3D (NMM3D). The real part of effective permittivity of SPF is also compared with solutions of Maxwell-Garnett formula.

In this paper, we develop a radar snow water equivalent (SWE) retrieval algorithm based on a parameterized forward model of bicontinuous dense media radiative transfer (Bic-DMRT). The algorithm is based on retrieving the absorption loss of the snowpack which is directly proportional to the SWE. In the algorithm, Bic-DMRT is first applied to generate a lookup table (LUT) of snowpack backscattering at X- and Ku-band. Regression training is applied to the LUT to transform the dual-frequency backscatter into functions of two parameters: the scattering albedo at X-band and SWE. The background scattering is subtracted from the SnowSAR data to give the volume scattering of snow. Classification of SnowSAR data is applied to provide a priori information. Based on the obtained volume scattering and the priori information, a cost function is established to find SWE. Performance of the retrieval algorithm was tested using three sets of airborne SnowSAR data acquired over mixed areas in Finland and open tundra landscape in Canada. It is shown that the retrieval algorithm has a root-mean-square error below 30 mm of SWE and a correlation coefficient above 0.64.

A fully coherent snowpack scattering and emission model is developed by numerically solving Maxwell's equations over the entire snowpack on a bottom half-space. The scattering matrix of the snowpack is directly obtained including both amplitude and phase. Both bistatic scattering coefficients and brightness temperatures of the snowpack are derived from full wave simulations. Simulation results demonstrate backscattering enhancement effects and coherent thin layer effects. The model is applied to study microwave signatures of the Arctic sea ice where the snow cover thickness has rapidly decreased. Microwave signatures are important in classification of sea-ice types and in quantitative characterization of snow cover properties. Both have strong impacts on the thermodynamics of sea ice. In the fully coherent model, a half-space dyadic Green's function is used in the volume integral equation to represent the effects of the underlying sea ice. Discrete dipole approximation is used to solve the volume integral equations, where parallel fast Fourier transform technique is utilized to accelerate the matrix-vector multiplications. The snowpack is represented as a bicontinuous medium. Periodic boundary conditions are applied in the two horizontal dimensions to simulate an infinite lateral extent of the snowpack.

We validate an absorption based radar retrieval algorithm of snow water equivalent (SWE) using X- and Ku-band backscatter with airborne SAR data. The bicontinuous dense media radiative transfer (Bic-DMRT) model is first applied to generate a look-up table of snow properties against backscattering at X- and Ku-bands. In the retrieval algorithm, the background scattering is subtracted from the total scattering giving the volume scattering of snow. With the look-up table, we generate regression equations between multiple and single scattering and correlations between the scattering albedo and optical thickness at the two bands. With these relationships and the volume scattering of the snowpack, the best solution for the radar observation is found using a priori constrained least-squares cost function. Next, the absorption loss of the snowpack is derived from the solution, which is directly proportional to the SWE. We have applied the algorithm to airborne SAR observations from Finland and Canada. The retrieval algorithm is shown to be effective, achieving root mean square error (RMSE) of ~19 mm for both SnowSAR data, which is smaller than the 20mm RMSE requirement of SCLP.