Literature-based automated discovery of tumor suppressor p53 phosphorylation and inhibition by NEK2.

Scientific progress depends on formulating testable hypotheses informed by the literature. In many domains, however, this model is strained because the number of research papers exceeds human readability. Here, we developed computational assistance to analyze the biomedical literature by reading PubMed abstracts to suggest new hypotheses. The approach was tested experimentally on the tumor suppressor p53 by ranking its most likely kinases, based on all available abstracts. Many of the best-ranked kinases were found to bind and phosphorylate p53 (P value = 0.005), suggesting six likely p53 kinases so far. One of these, NEK2, was studied in detail. A known mitosis promoter, NEK2 was shown to phosphorylate p53 at Ser315 in vitro and in vivo and to functionally inhibit p53. These bona fide validations of text-based predictions of p53 phosphorylation, and the discovery of an inhibitory p53 kinase of pharmaceutical interest, suggest that automated reasoning using a large body of literature can generate valuable molecular hypotheses and has the potential to accelerate scientific discovery.

COMPUTING THERAPY FOR PRECISION MEDICINE: COLLABORATIVE FILTERING INTEGRATES AND PREDICTS MULTI-ENTITY INTERACTIONS.

Biomedicine produces copious information it cannot fully exploit. Specifically, there is considerable need to integrate knowledge from disparate studies to discover connections across domains. Here, we used a Collaborative Filtering approach, inspired by online recommendation algorithms, in which non-negative matrix factorization (NMF) predicts interactions among chemicals, genes, and diseases only from pairwise information about their interactions. Our approach, applied to matrices derived from the Comparative Toxicogenomics Database, successfully recovered Chemical-Disease, Chemical-Gene, and Disease-Gene networks in 10-fold cross-validation experiments. Additionally, we could predict each of these interaction matrices from the other two. Integrating all three CTD interaction matrices with NMF led to good predictions of STRING, an independent, external network of protein-protein interactions. Finally, this approach could integrate the CTD and STRING interaction data to improve Chemical-Gene cross-validation performance significantly, and, in a time-stamped study, it predicted information added to CTD after a given date, using only data prior to that date. We conclude that collaborative filtering can integrate information across multiple types of biological entities, and that as a first step towards precision medicine it can compute drug repurposing hypotheses.

Supergenomic network compression and the discovery of EXP1 as a glutathione transferase inhibited by artesunate.

A central problem in biology is to identify gene function. One approach is to infer function in large supergenomic networks of interactions and ancestral relationships among genes; however, their analysis can be computationally prohibitive. We show here that these biological networks are compressible. They can be shrunk dramatically by eliminating redundant evolutionary relationships, and this process is efficient because in these networks the number of compressible elements rises linearly rather than exponentially as in other complex networks. Compression enables global network analysis to computationally harness hundreds of interconnected genomes and to produce functional predictions. As a demonstration, we show that the essential, but functionally uncharacterized Plasmodium falciparum antigen EXP1 is a membrane glutathione S-transferase. EXP1 efficiently degrades cytotoxic hematin, is potently inhibited by artesunate, and is associated with artesunate metabolism and susceptibility in drug-pressured malaria parasites. These data implicate EXP1 in the mode of action of a frontline antimalarial drug.

Persistent behavioral alterations in rats neonatally exposed to low doses of the organophosphate pesticide, parathion.

Although developmental exposures of rats to low levels of the organophosphate pesticides (OPs), chlorpyrifos (CPF) or diazinon (DZN), both cause persistent neurobehavioral effects, there are important differences in their neurotoxicity. The current study extended investigation to parathion (PTN), an OP that has higher systemic toxicity than either CPF or DZN. We gave PTN on postnatal days (PND) 1-4 at doses spanning the threshold for systemic toxicity (0, 0.1 or 0.2 mg/kg/day, s.c.) and performed a battery of emotional and cognitive behavioral tests in adolescence through adulthood. The higher PTN dose increased time spent on the open arms and the number of center crossings in the plus maze, indicating greater risk-taking and overall activity. This group also showed a decrease in tactile startle response without altering prepulse inhibition, indicating a blunted acute sensorimotor reaction without alteration in sensorimotor plasticity. T-maze spontaneous alternation, novelty-suppressed feeding, preference for sweetened chocolate milk, and locomotor activity were not significantly affected by neonatal PTN exposure. During radial-arm maze acquisition, rats given the lower PTN dose committed fewer errors compared to controls and displayed lower sensitivity to the amnestic effects of the NMDA receptor blocker, dizocilpine. No PTN effects were observed with regard to the sensitivity to blockade of muscarinic and nicotinic cholinergic receptors, or serotonin 5HT(2) receptors. This study shows that neonatal PTN exposure evokes long-term changes in behavior, but the effects are less severe, and in some incidences opposite in nature, to those seen earlier for CPF or DZN, findings consistent with our neurochemical studies showing different patterns of effects and less neurotoxic damage with PTN. Our results reinforce the conclusion that low dose exposure to different OPs can have quite different neurotoxic effects, obviously unconnected to their shared property as cholinesterase inhibitors.