Purpose: Ameloblastoma is an odontogenic neoplasm whose overall mutational landscape has not been well characterized. We sought to characterize pathogenic mutations in ameloblastoma and their clinical and functional significance with an emphasis on the mitogen-activated protein kinase (MAPK) pathway.Experimental Design: A total of 84 ameloblastomas and 40 non-ameloblastoma odontogenic tumors were evaluated with a combination of BRAF V600E allele-specific PCR, VE1 immunohistochemistry, the Ion AmpliSeq Cancer Hotspot Panel, and Sanger sequencing. Efficacy of a BRAF inhibitor was evaluated in an ameloblastoma-derived cell line.Results: Somatic, activating, and mutually exclusive RAS-BRAF and FGFR2 mutations were identified in 88% of cases. Somatic mutations in SMO, CTNNB1, PIK3CA, and SMARCB1 were also identified. BRAF V600E was the most common mutation, found in 62% of ameloblastomas and in ameloblastic fibromas/ fibrodentinomas but not in other odontogenic tumors. This mutation was associated with a younger age of onset, whereas BRAF wild-type cases arose more frequently in the maxilla and showed earlier recurrences. One hundred percent concordance was observed between VE1 immunohistochemistry and molecular detection of BRAF V600E mutations. Ameloblastoma cells demonstrated constitutive MAPK pathway activation in vitro. Proliferation and MAPK activation were potently inhibited by the BRAF inhibitor vemurafenib.Conclusions: Our findings suggest that activating FGFR2-RAS-BRAF mutations play a critical role in the pathogenesis of most cases of ameloblastoma. Somatic mutations in SMO, CTNNB1, PIK3CA, and SMARCB1 may function as secondary mutations. BRAF V600E mutations have both diagnostic and prognostic implications. In vitro response of ameloblastoma to a BRAF inhibitor suggests a potential role for targeted therapy. Clin Cancer Res; 20(21); 5517-26. Ó2014 AACR.

• Targeted

Inverted sinonasal papilloma (ISP) is a locally aggressive neoplasm associated with sinonasal squamous cell carcinoma (SNSCC) in 10-25% of cases. To date, no recurrent mutations have been identified in ISP or SNSCC. Using targeted next generation sequencing and Sanger sequencing, we identified activating EGFR mutations in 88% of ISP and 77% of ISP-associated SNSCC. Identical EGFR genotypes were found in matched pairs of ISP and associated SNSCC, providing the first genetic evidence of a biological link between these tumors. EGFR mutations were not identified in exophytic or oncocytic papillomas or non-ISP-associated SNSCC suggesting that the ISP/SNSCC spectrum is biologically distinct among sinonasal squamous tumors. Patients with ISP harboring EGFR mutations also exhibited an increased progression-free survival compared to those with wild-type EGFR. Finally, treatment of ISP-associated carcinoma cells with irreversible EGFR inhibitors resulted in inactivation of EGFR signaling and growth inhibition. These findings implicate a prominent role for activating EGFR mutations in the pathogenesis of ISP and associated SNSCC and rationalize consideration of irreversible EGFR inhibitors in the therapy of these tumors.

Large language models can encode a wealth of semantic knowledge about the world. Such knowledge could be extremely useful to robots aiming to act upon high-level, temporally extended instructions expressed in natural language. However, a significant weakness of language models is that they lack real-world experience, which makes it difficult to leverage them for decision making within a given embodiment. For example, asking a language model to describe how to clean a spill might result in a reasonable narrative, but it may not be applicable to a particular agent, such as a robot, that needs to perform this task in a particular environment. We propose to provide real-world grounding by means of pretrained skills, which are used to constrain the model to propose natural language actions that are both feasible and contextually appropriate. The robot can act as the language model's "hands and eyes," while the language model supplies high-level semantic knowledge about the task. We show how low-level skills can be combined with large language models so that the language model provides high-level knowledge about the procedures for performing complex and temporally extended instructions, while value functions associated with these skills provide the grounding necessary to connect this knowledge to a particular physical environment. We evaluate our method on a number of real-world robotic tasks, where we show the need for real-world grounding and that this approach is capable of completing long-horizon, abstract, natural language instructions on a mobile manipulator. The project's website and the video can be found at say-can.github.io. (a) Large Language Models (LLMs) (b) SayCan