Using Recurrent Neural Networks to Extract High-Quality Information From Lung Cancer Screening Computerized Tomography Reports for Inter-Radiologist Audit and Feedback Quality Improvement.

PURPOSE: Lung cancer screening programs generate a high volume of low-dose computed tomography (LDCT) reports that contain valuable information, typically in a free-text format. High-performance named-entity recognition (NER) models can extract relevant information from these reports automatically for inter-radiologist quality control. METHODS: Using LDCT report data from a longitudinal lung cancer screening program (8,305 reports; 3,124 participants; 2006-2019), we trained a rule-based model and two bidirectional long short-term memory (Bi-LSTM) NER neural network models to detect clinically relevant information from LDCT reports. Model performance was tested using F1 scores and compared with a published open-source radiology NER model (Stanza) in an independent evaluation set of 150 reports. The top performing model was applied to a data set of 6,948 reports for an inter-radiologist quality control assessment. RESULTS: The best performing model, a Bi-LSTM NER recurrent neural network model, had an overall F1 score of 0.950, which outperformed Stanza (F1 score = 0.872) and a rule-based NER model (F1 score = 0.809). Recall (sensitivity) for the best Bi-LSTM model ranged from 0.916 to 0.991 for different entity types; precision (positive predictive value) ranged from 0.892 to 0.997. Test performance remained stable across time periods. There was an average of a 2.86-fold difference in the number of identified entities between the most and the least detailed radiologists. CONCLUSION: We built an open-source Bi-LSTM NER model that outperformed other open-source or rule-based radiology NER models. This model can efficiently extract clinically relevant information from lung cancer screening computerized tomography reports with high accuracy, enabling efficient audit and feedback to improve quality of patient care.

A Non-Canonical Calmodulin Target Motif Comprising a Polybasic Region and Lipidated Terminal Residue Regulates Localization.

Calmodulin (CaM) is a Ca2+-sensor that regulates a wide variety of target proteins, many of which interact through short basic helical motifs bearing two hydrophobic 'anchor' residues. CaM comprises two globular lobes, each containing a pair of EF-hand Ca2+-binding motifs that form a Ca2+-induced hydrophobic pocket that binds an anchor residue. A central flexible linker allows CaM to accommodate diverse targets. Several reported CaM interactors lack these anchors but contain Lys/Arg-rich polybasic sequences adjacent to a lipidated N- or C-terminus. Ca2+-CaM binds the myristoylated N-terminus of CAP23/NAP22 with intimate interactions between the lipid and a surface comprised of the hydrophobic pockets of both lobes, while the basic residues make electrostatic interactions with the negatively charged surface of CaM. Ca2+-CaM binds farnesylcysteine, derived from the farnesylated polybasic C-terminus of KRAS4b, with the lipid inserted into the C-terminal lobe hydrophobic pocket. CaM sequestration of the KRAS4b farnesyl moiety disrupts KRAS4b membrane association and downstream signaling. Phosphorylation of basic regions of N-/C-terminal lipidated CaM targets can reduce affinity for both CaM and the membrane. Since both N-terminal myristoylated and C-terminal prenylated proteins use a Singly Lipidated Polybasic Terminus (SLIPT) for CaM binding, we propose these polybasic lipopeptide elements comprise a non-canonical CaM-binding motif.

Calmodulin disrupts plasma membrane localization of farnesylated KRAS4b by sequestering its lipid moiety.

KRAS4b is a small guanosine triphosphatase (GTPase) protein that regulates several signal transduction pathways that underlie cell proliferation, differentiation, and survival. KRAS4b function requires prenylation of its C terminus and recruitment to the plasma membrane, where KRAS4b activates effector proteins including the RAF family of kinases. The Ca2+-sensing protein calmodulin (CaM) has been suggested to regulate the localization of KRAS4b through direct, Ca2+-dependent interaction, but how CaM and KRAS4b functionally interact is controversial. Here, we determined a crystal structure, which was supported by solution nuclear magnetic resonance (NMR), that revealed the sequestration of the prenyl moiety of KRAS4b in the hydrophobic pocket of the C-terminal lobe of Ca2+-bound CaM. Our engineered fluorescence resonance energy transfer (FRET)-based biosensor probes (CaMeRAS) showed that, upon stimulation of Ca2+ influx by extracellular ligands, KRAS4b reversibly translocated in a Ca2+-CaM-dependent manner from the plasma membrane to the cytoplasm in live HeLa and HEK293 cells. These results reveal a mechanism underlying the inhibition of KRAS4b activity by Ca2+ signaling pathways.

Expression and Purification of Calmodulin for NMR and Other Biophysical Applications.

Calmodulin (CaM) is a ubiquitous calcium-sensing protein that has one of the most highly conserved sequences among eukaryotes. CaM has been a useful tool for biologists studying calcium signaling for decades. In recent years, CaM has also been implicated in numerous cancer-associated pathways, and rare CaM mutations have been identified as a cause of human cardiac arrhythmias. Here, we present a collection of our most recent and effective protocols for the expression and purification of recombinant CaM from Escherichia coli, including various isotopic labeling schemes, primarily for nuclear magnetic resonance (NMR) spectroscopy and other biophysical applications.

Tyrosyl phosphorylation of KRAS stalls GTPase cycle via alteration of switch I and II conformation.

Deregulation of the RAS GTPase cycle due to mutations in the three RAS genes is commonly associated with cancer development. Protein tyrosine phosphatase SHP2 promotes RAF-to-MAPK signaling pathway and is an essential factor in RAS-driven oncogenesis. Despite the emergence of SHP2 inhibitors for the treatment of cancers harbouring mutant KRAS, the mechanism underlying SHP2 activation of KRAS signaling remains unclear. Here we report tyrosyl-phosphorylation of endogenous RAS and demonstrate that KRAS phosphorylation via Src on Tyr32 and Tyr64 alters the conformation of switch I and II regions, which stalls multiple steps of the GTPase cycle and impairs binding to effectors. In contrast, SHP2 dephosphorylates KRAS, a process that is required to maintain dynamic canonical KRAS GTPase cycle. Notably, Src- and SHP2-mediated regulation of KRAS activity extends to oncogenic KRAS and the inhibition of SHP2 disrupts the phosphorylation cycle, shifting the equilibrium of the GTPase cycle towards the stalled 'dark state'.

The Novel ReNu Region of TAF12 Regulates Gcn5 Nucleosomal Acetylation.

The post-translational acetylation of the histone components of chromatin mediates numerous DNA-templated events, including transcriptional activation, DNA repair, and genomic replication. The conserved SAGA (Spt-Ada-Gcn5 Acetyltranferase) and SLIK (SAGA-Like) Histone Acetyltransferase (HAT) complexes are required for transcriptional activation of a subset of yeast genes and contain multiple subunits including the histone fold-containing TBP- Associated Factors (TAFs): 6, 9, 10, and 12. These TAFs are also components of the TFIID complex and are consequently involved in most RNA polymerase II-transcription in yeast. Here we identify a novel conserved region of TAF12, termed ReNu, outside of its histone fold, which is required for SAGA and SLIK-directed nucleosomal acetylation. We demonstrate that this region is not required for chromatin association, but show that this region plays an important role for histone H3 acetylation at specific SAGA and SLIK-regulated promoters. Our data suggests that the ReNu region of TAF12 regulates Gcn5 acetylation of specific substrates within the SAGA super-family of HAT complexes.