Comparison of the tumor immune microenvironment and checkpoint blockade biomarkers between stage III and IV non-small cell lung cancer.

BACKGROUND: Adjuvant immune checkpoint blockade (ICB) following chemoradiotherapy and adding ICB to chemotherapy have been key advances for stages III-IV non-small cell lung cancer (NSCLC) treatment. However, known biomarkers like PD-L1 are not consistently indicative of ICB response. Other markers within the tumor immune microenvironment (TIME) may better reflect ICB response and/or resistance mechanisms, but an understanding of how TIMEs differ between stage III and IV NSCLC has not been explored. METHODS: Real-world data from unresectable, stage III-IV, non-squamous, pretreatment NSCLCs (stage III n = 106, stage IV n = 285) were retrospectively analyzed. PD-L1 immunohistochemistry (IHC) was compared to CD274 gene expression. Then, differential gene expression levels, pathway enrichment, and immune infiltrate between stages were calculated from whole-transcriptome RNA-seq. Analyses were stratified by EGFR status. RESULTS: PD-L1 IHC and CD274 expression in tumor cells were highly correlated (n = 295, P < 2.2e-16, ⍴ = 0.74). CTLA4 expression was significantly increased in stage III tumors (P = 1.32e-04), while no differences were observed for other ICB-related genes. Metabolic pathway activity was significantly enriched in stage IV tumors (P = 0.004), whereas several immune-related KEGG pathways were enriched in stage III. Stage IV tumors had significantly increased macrophage infiltration (P = 0.0214), and stage III tumors had a significantly higher proportion of CD4 + T cells (P = 0.017). CD4 + T cells were also relatively more abundant in EGFR-mutant tumors vs. wild-type (P = 0.0081). CONCLUSION: Directly comparing the TIMEs of stage III and IV NSCLC, these results carry implications for further studies of ICB response in non-resectable stage III NSCLC and guide further research of prognostic biomarkers and therapeutic targets.

Riboflow: Using Deep Learning to Classify Riboswitches With ∼99% Accuracy.

Riboswitches are cis-regulatory genetic elements that use an aptamer to control gene expression. Specificity to cognate ligand and diversity of such ligands have expanded the functional repertoire of riboswitches to mediate mounting apt responses to sudden metabolic demands and signal changes in environmental conditions. Given their critical role in microbial life, riboswitch characterisation remains a challenging computational problem. Here we have addressed the issue with advanced deep learning frameworks, namely convolutional neural networks (CNN), and bidirectional recurrent neural networks (RNN) with Long Short-Term Memory (LSTM). Using a comprehensive dataset of 32 ligand classes and a stratified train-validate-test approach, we demonstrated the accurate performance of both the deep learning models (CNN and RNN) relative to conventional hyperparameter-optimized machine learning classifiers on all key performance metrics, including the ROC curve analysis. In particular, the bidirectional LSTM RNN emerged as the best-performing learning method for identifying the ligand-specificity of riboswitches with an accuracy >0.99 and macro-averaged F-score of 0.96. An additional attraction is that the deep learning models do not require prior feature engineering. A dynamic update functionality is built into the models to factor for the constant discovery of new riboswitches, and extend the predictive modeling to new classes. Our work would enable the design of genetic circuits with custom-tuned riboswitch aptamers that would effect precise translational control in synthetic biology. The associated software is available as an open-source Python package and standalone resource for use in genome annotation, synthetic biology, and biotechnology workflows.

PromoterPredict: sequence-based modelling of Escherichia coli σ70 promoter strength yields logarithmic dependence between promoter strength and sequence.

We present PromoterPredict, a dynamic multiple regression approach to predict the strength of Escherichia coli promoters binding the σ70 factor of RNA polymerase. σ70 promoters are ubiquitously used in recombinant DNA technology, but characterizing their strength is demanding in terms of both time and money. We parsed a comprehensive database of bacterial promoters for the -35 and -10 hexamer regions of σ70-binding promoters and used these sequences to construct the respective position weight matrices (PWM). Next we used a well-characterized set of promoters to train a multivariate linear regression model and learn the mapping between PWM scores of the -35 and -10 hexamers and the promoter strength. We found that the log of the promoter strength is significantly linearly associated with a weighted sum of the -10 and -35 sequence profile scores. We applied our model to 100 sets of 100 randomly generated promoter sequences to generate a sampling distribution of mean strengths of random promoter sequences and obtained a mean of 6E-4 ± 1E-7. Our model was further validated by cross-validation and on independent datasets of characterized promoters. PromoterPredict accepts -10 and -35 hexamer sequences and returns the predicted promoter strength. It is capable of dynamic learning from user-supplied data to refine the model construction and yield more robust estimates of promoter strength. PromoterPredict is available as both a web service (https://promoterpredict.com) and standalone tool (https://github.com/PromoterPredict). Our work presents an intuitive generalization applicable to modelling the strength of other promoter classes.