Accurate prediction of antigen presentation by human leukocyte antigen (HLA) class II molecules would be valuable for vaccine development and cancer immunotherapies. Current computational methods trained on in vitro binding data are limited by insufficient training data and algorithmic constraints. Here we describe MARIA (major histocompatibility complex analysis with recurrent integrated architecture; https://maria.stanford.edu/), a multimodal recurrent neural network for predicting the likelihood of antigen presentation from a gene of interest in the context of specific HLA class II alleles. In addition to in vitro binding measurements, MARIA is trained on peptide HLA ligand sequences identified by mass spectrometry, expression levels of antigen genes and protease cleavage signatures. Because it leverages these diverse training data and our improved machine learning framework, MARIA (area under the curve = 0.89-0.92) outperformed existing methods in validation datasets. Across independent cancer neoantigen studies, peptides with high MARIA scores are more likely to elicit strong CD4 + T cell responses. MARIA allows identification of immunogenic epitopes in diverse cancers and autoimmune disease.

The phosphatidylinositol phosphate kinase (PIPK) family of enzymes is primarily responsible for converting singly phosphorylated phosphatidylinositol derivatives to phosphatidylinositol bisphosphates. As such, these kinases are central to many signaling and membrane trafficking processes in the eukaryotic cell. The three types of phosphatidylinositol phosphate kinases are homologous in sequence but differ in catalytic activities and biological functions. Type I and type II kinases generate phosphatidylinositol 4,5-bisphosphate from phosphatidylinositol 4-phosphate and phosphatidylinositol 5-phosphate, respectively, whereas the type III kinase produces phosphatidylinositol 3,5-bisphosphate from phosphatidylinositol 3-phosphate. Based on crystallographic analysis of the zebrafish type I kinase PIP5K α , we identified a structural motif unique to the kinase family that serves to recognize the monophosphate on the substrate. Our data indicate that the complex pattern of substrate recognition and phosphorylation results from the interplay between the monophosphate binding site and the specificity loop: the specificity loop functions to recognize different orientations of the inositol ring, whereas residues flanking the phosphate binding Arg244 determine whether phosphatidylinositol 3-phosphate is exclusively bound and phosphorylated at the 5-position. This work provides a thorough picture of how PIPKs achieve their exquisite substrate specificity.lipid kinases | protein engineering | crystallography | substrate specificity P hosphatidylinositol phosphate (PIP) kinases (PIPKs) are key players in the metabolism of phosphoinositides in eukaryotic cells. PIPKs are primarily responsible for the synthesis of doubly phosphorylated phosphatidylinositol (PI) derivatives from singly phosphorylated PIs (1-4). PIPK catalytic activities are important to the cell in part because they produce essential PI bisphosphates such as PI(4,5)P 2 , which is involved in a wide variety of signaling pathways and membrane trafficking events, and PI(3,5)P 2 , which has a more specialized role in endosomes (5). In addition, these kinases control the level of certain PI monophosphates such as PI(5)P, which appears to function as a stress signal (6). Mutations in PIPKs have been linked to human diseases (7,8), and, in cancerous cells, the activities of PIPKs are often up-regulated, usually as a consequence of overexpression (9, 10).There are three subfamilies of PIPKs that share sequence identity within the kinase domain (2). The type I kinase phosphatidylinositol 4-phosphate 5-kinase (PIP5K), localized mainly to the plasma membrane, is responsible for synthesizing the majority of PI(4,5)P 2 in the cell and accomplishes this process by phosphorylating the 5-hydroxyl group of PI(4)P (Fig. S1A). In vitro, the type I kinase also has a robust activity against PI(3)P, generating not only PI(3,4)P 2 and PI(3,5)P 2 but also the triply phosphorylated PI(3,4,5)P 3 (11, 12). Furthermore, the type I kinase can weakly phosphorylate PI to produce PI(5)P. In con...

OBJECTIVEStereoelectroencephalography (SEEG) has increased in popularity for localization of epileptogenic zones in drug-resistant epilepsy because safety, accuracy, and efficacy have been well established in both adult and pediatric populations. Development of robot-guidance technology has greatly enhanced the efficiency of this procedure, without sacrificing safety or precision. To date there have been very limited reports of the use of this new technology in children. The authors present their initial experience using the ROSA platform for robot-guided SEEG in a pediatric population.METHODSBetween February 2016 and October 2017, 20 consecutive patients underwent robot-guided SEEG with the ROSA robotic guidance platform as part of ongoing seizure localization and workup for medically refractory epilepsy of several different etiologies. Medical and surgical history, imaging and trajectory plans, as well as operative records were analyzed retrospectively for surgical accuracy, efficiency, safety, and epilepsy outcomes.RESULTSA total of 222 leads were placed in 20 patients, with an average of 11.1 leads per patient. The mean total case time (± SD) was 297.95 (± 52.96) minutes and the mean operating time per lead was 10.98 minutes/lead, with improvements in total (33.36 minutes/lead vs 21.76 minutes/lead) and operative (13.84 minutes/lead vs 7.06 minutes/lead) case times/lead over the course of the study. The mean radial error was 1.75 (± 0.94 mm). Clinically useful data were obtained from SEEG in 95% of cases, and epilepsy surgery was indicated and performed in 95% of patients. In patients who underwent definitive epilepsy surgery with at least a 3-month follow-up, 50% achieved an Engel class I result (seizure freedom). There were no postoperative complications associated with SEEG placement and monitoring.CONCLUSIONSIn this study, the authors demonstrate that rapid adoption of robot-guided SEEG is possible even at a SEEG-naïve institution, with minimal learning curve. Use of robot guidance for SEEG can lead to significantly decreased operating times while maintaining safety, the overall goals of identification of epileptogenic zones, and improved epilepsy outcomes.

No abstract

No abstract