Neural methods of molecule property prediction require efficient encoding of structure and property relationship to be accurate. Recent work using graph algorithms shows limited generalization in the latent molecule encoding space. We build a Transformer-based molecule encoder and property predictor network with novel input featurization that performs significantly better than existing methods. We adapt our model to semi-supervised learning to further perform well on the limited experimental data usually available in practice.
Andrew Silberfarb
SRI Author
Deep Adaptive Semantic Logic (DASL): Compiling Declarative Knowledge into Deep Neural Networks
We introduce Deep Adaptive Semantic Logic (DASL), a novel framework for automating the generation of deep neural networks that incorporates user-provided formal knowledge to improve learning from data. We provide formal semantics that demonstrate that our knowledge representation captures all of first order logic and that finite sampling from infinite domains converges to correct truth values. DASL’s representation improves on prior neural-symbolic work by avoiding vanishing gradients, allowing deeper logical structure, and enabling richer interactions between the knowledge and learning components. We illustrate DASL through a toy problem in which we add structure to an image classification problem and demonstrate that knowledge of that structure reduces data requirements by a factor of 1000 . We then evaluate DASL on a visual relationship detection task and demonstrate that the addition of commonsense knowledge improves performance by 10.7 % in a data scarce setting.
Resources to discover and use short linear motifs in viral proteins
Viral proteins evade host immune function by molecular mimicry, often achieved by short linear motifs (SLiMs) of three to ten consecutive amino acids (AAs). Motif mimicry tolerates mutations, evolves quickly to modify interactions with the host, and enables modular interactions with protein complexes. Host cells cannot easily coordinate changes to conserved motif recognition and binding interfaces under selective pressure to maintain critical signaling pathways. SLiMs offer potential for use in synthetic biology, such as better immunogens and therapies, but may also present biosecurity challenges. We survey viral uses of SLiMs to mimic host proteins, and information resources available for motif discovery. As the number of examples continues to grow, knowledge management tools are essential to help organize and compare new findings.
