GlioPredictor: A Deep Learning Model for Identification of High-Risk Low-Grade Glioma toward Adjuvant Treatment Planning

PURPOSE/OBJECTIVE(S)

High-risk low-grade glioma (LGG) patients are recommended to undergo adjuvant radiotherapy whereas watchful waiting is recommended for low-risk LGG patients per the latest NCCN guidelines. Based on Radiation Therapy Oncology Group (RTOG) 9802, high-risk features include age >40 or subtotal resection (STR). However, in the era of molecular-based classification for tumors of central neural system, current risk classification criteria based on gross disease and patient demographics may be outdated. Here, we aim to develop a molecular-based glioma risk classification system (GlioPredictor) that could potentially facilitate identification of high-risk LGG patients.

MATERIALS/METHODS

A total of 507 LGG cases from The Cancer Genome Atlas-low grade glioma (TCGA-LGG), and 1,309 cases from AACR GENIE v13.0 datasets were studied for genetic disparities between IDH1-wildtype and mutated cohorts, and varying age groups. Through a feature selection technique using genomic profiling and correlation analyses, features such as mutation status, copy number variations (CNVs), among other clinicopathologic features prognostic of IDH1 mutation status were selected as potential inputs to train an artificial neural networks (ANNs) that could predict IDH1 mutation status. Model performance was assessed using the area under the receiver operating characteristic curve (AUC). Memorial Sloan Kettering (MSK) dataset (n = 404) for LGG was used to cross-validate the trained ANN. The optimized ANN model has 6 layers with 6 input nodes, 20 hidden nodes, and a binary output layer. The weights and biases of the hidden layers of the best-performing model were retrieved and reconstructed to yield the GlioPredictor score-the predicted risk of progression for IDH1-wildtype LGG.

RESULTS

Over 81% of glioma patients age less than 40 have IDH1 mutation, as compared with 31% in those age above 60. Using age > 40 as a cutoff failed to identify high-risk IDH1-mutant LGG with early progression. IDH1 mutation is associated with decreased CNVs of EGFR (21 % vs. 3%), CDKN2A (20% vs. 6%) and PTEN (14% vs. 1.7%), and increased percentage of mutations for TP53 (15% vs. 63%), and ATRX (10% vs. 54%) (p<0.001). Using these molecular features, along with the patient's age, an ANN model with 6 layers and 20 hidden nodes can predict IDH1 mutation status with over 90% accuracy and AUC score over 0.91.

CONCLUSION

We have developed an ANN model that is capable of learning the prognostic features of LGG associated with an IDH1-mutated LGG cohort and using the features to predict high-risk patients from the IDH1-wildtype cohort. This ANN model facilitates the selection of LGG patients who could benefit from immediate adjuvant radiotherapy. Future work includes the integration of image features to improve the prediction performance of the GlioPredictor system.

Cost Analysis of MR-Guided vs. CT-Guided Radiation Therapy for Locally Advanced Pancreatic Cancer

PURPOSE/OBJECTIVE(S)

Stereotactic magnetic resonance guided on-table adaptive radiotherapy (SMART) is an increasingly utilized radiotherapy (RT) treatment for locally advanced pancreatic cancer (LAPC), providing improved local control and overall survival with reduced toxicity. Computed tomography (CT) guided RT options include stereotactic body radiotherapy (SBRT) and hypofractionated RT with volumetric modulated arc therapy (VMAT). Currently there are no financial comparisons for MR vs CT-guided RT for LAPC. We completed a cost analysis comparing the physician and hospital charges associated with RT options for LAPC.

MATERIALS/METHODS

To compare RT costs, we identified Current Procedural Terminology (CPT) codes utilized for 5-fraction SMART with adaptation (50 Gy, biological effective dose (BED) 100 Gy), 5-fraction CT-guided SBRT (33 Gy, BED 55 Gy), and 15-fraction CT-guided hypofractionated VMAT (67.5 Gy, BED 98 Gy) in a hospital-based practice setting. Physician and hospital Medicare prices associated with these codes together summarize the overall cost. We determined physician fees using the Centers for Medicare and Medicaid Services (CMS) Physician Fee Schedule Search to search the Healthcare Common Procedure Coding System (HCPCS) for "professional" costs included within "facility" costs. We determined hospital fees using the Outpatient Prospective Payment System addendum. To standardize costs, we searched for national payment amounts for the 2022 calendar year.

RESULTS

Total cost of SMART with adaptation was 136% higher than the cost of CT-SBRT and 149% higher than the cost of hypofractionated RT. Physician fees for SMART were 173% higher and 157% higher than the fees for CT-SBRT and hypofractionated RT, respectively. Hospital fees for SMART were 129% higher and 147% higher than the fees for CT-SBRT and hypofractionated RT, respectively. The total cost of CT-SBRT was only 5% higher than cost of hypofractionated RT. The physician fees for hypofractionated RT were 6% higher than those for CT-SBRT, while the outpatient fees for CT SBRT were 7% higher than those for hypofractionated RT.

CONCLUSION

With recent data demonstrating favorable efficacy and toxicity rates for SMART, practices may increasingly consider investing in this treatment modality. This is the first cost analysis comparing SMART to CT-guided SBRT and hypofractionated RT in LAPC. We demonstrate higher costs of SMART compared to CT-guided RT, attributable primarily to higher number of dosimetry calculations for this modality and for adapted fractions. We also demonstrate comparable costs of lower BED CT-guided SBRT and higher BED hypofractionated RT. Further investigation is needed to assess whether the survival benefit of SMART translates to favorable cost per quality adjusted life year.

Stage IIIA Non-Small Cell Lung Cancer Treatment and Outcomes: A Single Institution Retrospective Analysis

PURPOSE/OBJECTIVE(S)

Treatment of locally advanced non-small cell lung cancer (NSCLC) remains challenging, with a multitude of treatment options available for Stage III patients. We hypothesized that Stage IIIA outcomes differ by treatment received.

MATERIALS/METHODS

We performed a retrospective review of NSCLC patients ≥18 years old with Stage IIIA disease treated 1/1/2010-03/01/2022. Demographics, treatment received, treatment outcomes, and failure patterns were collected. Progression-free survival (PFS) and overall survival (OS) were assessed using Kaplan-Meier analysis. Kruskal-Wallis ANOVA was used to compare groups.

RESULTS

Of 352 patients identified, 160 had Stage IIIA NSCLC with a median follow-up of 29.1 months. Patients had a median age of 63 years, 79 (49.4%) were male, and 137 (85.6%) were current/former smokers (with 30 median pack-years). Patients were treated as follows: 17 (11%) surgery alone (S), 91 (57%) definitive radiation ± chemotherapy (CRT), 52 (33%) neoadjuvant therapy followed by surgery (Neo). 6 (12%) of the Neo group received chemoimmunotherapy, and 21 (51%) of the 41 CRT patients received adjuvant immunotherapy. Between the three groups, there were no significant differences in tumor size as measured by T-staging (p = 0.83) and baseline FEV1/FVC (p = 0.92). Median PFS was 33.5mo (95% CI 13.2-NA) for group S, 18.4mo (95% CI 12.7-42.2) for CRT, and 19.7mo (95% CI 13.9-NA) for Neo with no significant intergroup difference (p = 0.72). Median OS was 33.5mo (95% CI 13.2-NA) for S, 48.7mo (95% CI 36.0-88.9) for CRT, and 50.9mo (95% CI 41.9-NA) for Neo with no significant intergroup difference (p = 0.94). Among the 17 primary surgical patients, 11 (65%) experienced failure: 6 (35%) local, 5 (29%) regional, and 7 (41%) distant. Among the 91 CRT patients, 57 (63%) experienced failure: 40 (44%) local, 35 (38%) regional, and 28 (31%) distant. Among the 52 Neo patients, 26 (50%) experienced failure: 14 (27%) local, 15 (29%) regional, and 17 (33%) distant. There were no significant differences in rates of local failure (p = 0.26), regional failure (p = 0.59), distant failure (p = 0.79), or any failure (p = 0.41) among the three treatment groups. The most common locations for distant failure were pleural effusions (n = 15, 29%), CNS (n = 14, 27%), and bone (n = 11, 21%).

CONCLUSION

In this single institution retrospective study, we find no significant differences in PFS, OS, and failure patterns between patients with Stage IIIA NSCLC treated with definitive (chemo)radiation and neoadjuvant therapy. Numeric improvement in PFS in surgery-only patients is consistent with expected patient selection of this group. Further work in the immunotherapy era is needed.

Role of Radiation Therapy in Liver-Only Oligometastatic Disease: A SEER Analysis

PURPOSE/OBJECTIVE(S)

Radiation therapy (RT) for oligometastasis has the potential to prolong survival in certain disease sites. There is a paucity of data regarding the benefit of RT for overall survival (OS) or disease-specific survival (DSS) in patients with liver-only oligometastatic disease.

MATERIALS/METHODS

The Surveillance, Epidemiology and End Results Program (SEER) includes comprehensive metastasis data for patients from 2016-2019. The SEER database was queried for patients with liver-only metastatic disease at diagnosis by selecting stage IV cases with liver-only metastasis, without metastatic disease in bone, brain, lung, distant lymph nodes, or other sites. OS and DSS were estimated using Kaplan-Meier with log-rank analysis to compare patients who received RT versus no RT. Cox proportional hazards regression was applied to identify potential confounders. Subgroup analysis was used to explore the benefit of RT in different primary tumor sites including pancreas (N = 8846), followed by colon (N = 6535), lung (N = 3075), rectum (N = 1739), and stomach (N = 1448).

RESULTS

A total of 29,422 patients with liver-only metastatic disease treated from 2016-2019 were included. The median age was 67 years old and 77.0% of the patients were Caucasian. 2448 (8.3%) patients were confirmed to have received RT. Patients who received RT had better OS (median survival, RT vs no RT: 18 vs 6 months, P<0.001) and DSS (18 vs 7 months, P<0.001). On multivariable analyses, RT still significantly improved both OS (HR: 0.705, 95% CI: 0.665-0.747, P<0.001) and DSS (HR: 0.390, 95% CI: 0.378-0.402, P<0.001) after adjusting for potential confounders, including age, tumor size, lymph node status, and chemotherapy. RT was significantly associated with improved OS and DSS (all P<0.001) in all primary tumors sites queried except for stomach primary for which RT did not impact OS (P = 0.122) and DSS (P = 0.229). In patients who received chemotherapy, RT also prolonged OS (P<0.001) and DSS (P<0.001).

CONCLUSION

In the SEER database of patients with liver-only oligometastatic disease, RT improves OS as well as DSS, however the benefit varies for the different primary tumor sites. Prospective studies could help further clarify the survival benefits of RT in liver-only oligometastatic disease.

An intramolecular contact in Galpha transducin that participates in maintaining its intrinsic GDP release rate

Receptor mediated stimulation of the G protein-alpha subunit leads to exchange of GDP for GTP, activating the protein. Spontaneous GDP release from Galpha can also lead to the active state, if GTP in solution binds the nucleotide binding pocket. The purpose of this study is to evaluate the molecular determinants for maintaining the spontaneous GDP release rates between two Galpha subunits. Galpha(t) has a low rate of nucleotide release, compared to Galpha(i1). Galpha(t/i1) chimeras were used to explore the molecular basis for this behavior. The C-terminal alpha4-helix, the N-terminal 56 residues and the Switch I/II regions of Galpha(t) were shown to affect the low spontaneous GDP release rate in Galpha(t). A specific molecular contact between Asp26 and Asn191 was found in Galpha(t) that is not present in Galpha(i1). In two chimeras disrupting this interaction produced an increased spontaneous GDP release; restoring the contact present in Galpha(t) into these chimeras decreased the GDP release rate by half as compared to the original chimeras. Similarly, introduction of this contact in wild-type Galpha(i1) decreased the GDP release rate of Galpha(i1) by half. Differences in GDP release rates may reflect physiological roles these proteins play in living systems.

G alpha minigenes expressing C-terminal peptides serve as specific inhibitors of thrombin-mediated endothelial activation

The C termini of G protein alpha subunits are critical for binding to their cognate receptors, and peptides corresponding to the C terminus can serve as competitive inhibitors of G protein-coupled receptor-G protein interactions. This interface is quite specific as a single amino acid difference annuls the ability of a G alpha(i) peptide to bind the A(1) adenosine receptor (Gilchrist, A., Mazzoni, M., Dineen, B., Dice, A., Linden, J., Dunwiddie, T., and Hamm, H. E. (1998 ) J. Biol. Chem. 273, 14912--14919). Recently, we demonstrated that a plasmid minigene vector encoding the C-terminal sequence of G alpha(i) could specifically inhibit downstream responses to agonist stimulation of the muscarinic M(2) receptor (Gilchrist, A., Bunemann, M., Li, A., Hosey, M. M., and H. E. Hamm (1999) J. Biol. Chem. 274, 6610--6616). To selectively antagonize G protein signal transduction events and determine which G protein underlies a given thrombin-induced response, we generated minigene vectors that encode the C-terminal sequence for each family of G alpha subunits. Minigene vectors expressing G alpha C-terminal peptides (G alpha(i), G alpha(q), G alpha(12), and G alpha(13)) or the control minigene vector, which expresses the G alpha(i) peptide in random order (G(iR)), were systematically introduced into a human microvascular endothelial cell line. The C-terminal peptides serve as competitive inhibitors presumably by blocking the site on the G protein-coupled receptor that normally binds the G protein. Our results not only confirm that each G protein can control certain signaling events, they emphasize the specificity of the G protein-coupled receptor-G protein interface. In addition, the C-terminal G alpha minigenes appear to be a powerful tool for dissecting out the G protein that mediates a given physiological function following thrombin activation.

Tissue-specific expression of a Ca(2+)-activated K+ channel is controlled by multiple upstream regulatory elements

The electrical properties of a cell are produced by the complement of ion channels that it expresses. To understand how ion-channel gene expression is regulated, we are studying the tissue-specific regulation of the slowpoke (slo) Ca(2+)-activated K+ channel gene. This gene is expressed in the central and peripheral nervous system, in midgut and tracheal cells, and in the musculature of Drosophila melanogaster. The entire transcriptional control region has been cloned previously and shown to reproduce the tissue and developmental expression pattern of the endogenous gene. Here we demonstrate that s/o has at least four promoters distributed over approximately 4.5 kb of DNA. Promoter C1 and C1c display a TATA box-like sequence at the appropriate distance from the transcription start site. Promoters C1b and C2, however, are TATA-less promoters. C1, C1b, and C1c transcripts differ in their leader sequence but share a common translation start site. C2 transcripts incorporate a new translation start site that appends 17 amino acids to the N terminus of the encoded protein. Deletion analysis was used to identify sequences important for tissue-specific expression. We used a transgenic in vivo expression system in which all tissues and developmental stages can be assayed easily. Six nested deletions were transformed into Drosophila, and the expression pattern was determined using a lacZ reporter in both dissected tissues and sectioned animals. We have identified different sequences required for expression in the CNS, midgut, tracheal cells, and muscle.