The Power of Translator
Generating new hypotheses for potential investigation through clinical trials
Translator identified several potential treatments for an unknown respiratory disorder based on a known genetic mutation. Working closely with
clinicians, an option with relatively low risk was identified and when tried, significantly increased airway function.
Process Details
Translator was queried to find drugs that could reduce the PRDX1 and TLR4 proteins. One of Translator’s reasoning tools, mediKanren, returned several potential treatments based on database assertions from the Semantic MEDLINE
Datbase (SemMedDB). These included dioscin, a steroid that induces DNA damage and cell death and downregulates PRDX1, and naltrexone, an opioid agonist that blocks the action of TLR4. These two drugs were returned to the
patient’s physician as a potential treatment option, but the physician was concerned about potential side effects from these two drugs, and asked the researcher to continue the search.
Since potential treatments to directly target PRDX1 and TLR4 were too risky, upstream regulators of these genes were investigated next. Translator was queried to find genes that regulate PRDX1, and erythroid-like nuclear factor 2 (NFE2L2) was returned, which is known to upregulate PRDX1. Reduction of NFE2L2 therefore can be assumed to also reduce PRDX1, so Translator was then queried to find drugs that can reduce NFE2L2, and a pair of
dietary supplements, ascorbic acid and luteolin, were returned based on SemMedDB assertions. Since dietary supplements are not likely to cause major side effects, these supplements were returned to the patient’s clinician, who
decided to recommend that the patient try the supplements. Three months after taking the supplements, the patient’s airway function had improved significantly, by 41%.
Translator in the Media
How an AI expert took on his toughest project ever: writing code to save his son’s life
Data Sources Connected
RTX
A reasoning tool created for Translator that integrates many different databases into a cohesive knowledge graph.
SemMedDB
A repository of semantic predications (subject-predicate-object triples) extracted by SemRep, a semantic interpreter of biomedical text. SemMedDB currently contains approximately 94.0 million predications extracted from
PubMed.
MONARCH
A database linking model organism research to human disease allowing for the comparison of similar phenotypic profiles across species.
mediKanren
A reasoning tool created for Translator that can access knowledge graphs to find novel connections between data points.
References
https://www.nature.com/articles/s41598-017-06484-6
https://skr3.nlm.nih.gov/SemMedDB/dbinfo.html
https://monarchinitiative.org/