On the path to full understanding of the structure-function relationship or even design of RNA, structure prediction would offer an intriguing complement to experimental efforts. Any deep learning on RNA structure, however, is hampered by the sparsity of labeled training data. Utilizing the limited data available, we here focus on predicting spatial adjacencies (”contact maps”) as a proxy for 3D structure. We explore the space of self-supervised learning for RNA multiple sequence alignments and focus on downstream contact prediction from latent attention maps. Boosted decision trees in particular advance contact prediction quality and can be further enhanced by finetuning the pretrained backbone. Impressively, they double the precision of contact prediction/ reduce false positives by a factor of five over the baseline. We name our model BARNACLE. Our conceptional advance could prove a breakthrough in decreasing the sequence-structure gap for RNA and is generalize-able to other tasks.

With increasing data and model complexities, the time required to train neural networks has become prohibitively large. To address the exponential rise in training time, users are turning to data parallel neural networks (DPNN) and large-scale distributed resources on computer clusters. Current DPNN approaches implement the network parameter updates by synchronizing and averaging gradients across all processes with blocking communication operations after each forward-backward pass. This synchronization is the central algorithmic bottleneck. We introduce the distributed asynchronous and selective optimization (DASO) method, which leverages multi-GPU compute node architectures to accelerate network training while maintaining accuracy. DASO uses a hierarchical and asynchronous communication scheme comprised of node-local and global networks while adjusting the global synchronization rate during the learning process. We show that DASO yields a reduction in training time of up to 34% on classical and state-of-the-art networks, as compared to current optimized data parallel training methods.

With increasing data and model complexities, the time required to train neural networks has become prohibitively large. To address the exponential rise in training time, users are turning to data parallel neural networks (DPNN) and large-scale distributed resources on computer clusters. Current DPNN approaches implement the network parameter updates by synchronizing and averaging gradients across all processes with blocking communication operations after each forward-backward pass. This synchronization is the central algorithmic bottleneck. We introduce the Distributed Asynchronous and Selective Optimization (DASO) method, which leverages multi-GPU compute node architectures to accelerate network training while maintaining accuracy. DASO uses a hierarchical and asynchronous communication scheme comprised of node-local and global networks while adjusting the global synchronization rate during the learning process. We show that DASO yields a reduction in training time of up to 34% on classical and state-of-the-art networks, as compared to current optimized data parallel training methods.