Probabilistic Deterministic Finite Automata and Recurrent Networks, Revisited

Reservoir computers (RCs) and recurrent neural networks (RNNs) can mimic any finite-state automaton in theory, and some workers demonstrated that this can hold in practice. We test the capability of generalized linear models, RCs, and Long Short-Term Memory (LSTM) RNN architectures to predict the stochastic processes generated by a large suite of probabilistic deterministic finite-state automata (PDFA) in the small-data limit according to two metrics: predictive accuracy and distance to a predictive rate-distortion curve. The latter provides a sense of whether or not the RNN is a lossy predictive feature extractor in the information-theoretic sense. PDFAs provide an excellent performance benchmark in that they can be systematically enumerated, the randomness and correlation structure of their generated processes are exactly known, and their optimal memory-limited predictors are easily computed. With less data than is needed to make a good prediction, LSTMs surprisingly lose at predictive accuracy, but win at lossy predictive feature extraction. These results highlight the utility of causal states in understanding the capabilities of RNNs to predict.

Novel Development of Predictive Feature Fingerprints to Identify Chemistry-Based Features for Effective Drug Design of SARS-CoV-2 Target Antagonists and Inhibitors Using Machine Learning

<p>A unique approach to bioactivity and chemical data curation coupled with Random forest analyses has led to a series of target-specific and cross-validated Predictive Feature Fingerprints (PFF) that have high predictability across multiple therapeutic targets and disease stages involved in the SARS-CoV-2 induced COVID-19 pandemic, which include plasma kallikrein, HIV protease, NSP5, NSP12, JAK family and AT-1. The approach was highly accurate in determining the matched target for the different compound sets and suggests that the models could be used for virtual screening of target specific compound libraries. The curation-modeling process was successfully applied to a SARS-CoV-2 phenotypic screen and could be used for predictive bioactivity estimation and prioritization for clinical trial selection, virtual screening of drug libraries for repurposing of drug molecules, and analysis and direction of proprietary datasets.</p>

Novel Development of Predictive Feature Fingerprints to Identify Chemistry-Based Features for Effective Drug Design of SARS-CoV-2 Target Antagonists and Inhibitors Using Machine Learning

<p>A unique approach to bioactivity and chemical data curation coupled with Random forest analyses has led to a series of target-specific and cross-validated Predictive Feature Fingerprints (PFF) that have high predictability across multiple therapeutic targets and disease stages involved in the SARS-CoV-2 induced COVID-19 pandemic, which include plasma kallikrein, HIV protease, NSP5, NSP12, JAK family and AT-1. The approach was highly accurate in determining the matched target for the different compound sets and suggests that the models could be used for virtual screening of target specific compound libraries. The curation-modeling process was successfully applied to a SARS-CoV-2 phenotypic screen and could be used for predictive bioactivity estimation and prioritization for clinical trial selection, virtual screening of drug libraries for repurposing of drug molecules, and analysis and direction of proprietary datasets.</p>

Drug Combination Antagonism and Single Agent Dominance Result from Differences in Death Activation Kinetics

ABSTRACTTherapeutic regimens for cancer generally involve drugs used in combinations. Most prior work has focused on identifying and understanding synergistic drug-drug interactions; however, understanding sources of antagonistic interactions remains an important and understudied issue. To enrich for antagonistic interactions and reveal common features of these drug combinations, we screened all pairwise combinations of drugs characterized as canonical activators of different forms of regulated cell death. We find that this network is strongly enriched for antagonistic interactions, and in particular, enriched for an extreme form of antagonism, which we call “single agent dominance”. Single agent dominance refers to antagonisms in which a two drug combination phenocopies one of the two agents. We find that dominance results from differences in the cell death onset time, with dominant drugs inducing death earlier and at faster rates than their suppressed counterparts. Finally, we explored the mechanisms by which parthanatotic agents dominate apoptotic agents, finding that dominance in this scenario is caused by mutually exclusive and conflicting use of PARP1. Taken together, our study reveals death activation kinetics as a predictive feature of antagonism, due to inhibitory crosstalk between cell death pathways.