Highlighted Projects

Key Findings:

• Platelet-to-T cell ratio (Pla-T ratio) is a strong predictor of patient survival.
• Platelets play a role in lymphocyte activation, proliferation, and differentiation.
• Platelet-monocyte aggregation is associated with disease severity in both COVID-19 and sepsis.
• We identified distinct platelet subpopulations that were overrepresented in convalescent, surviving, and fatal patients.
• Notably, three types of platelets were highly represented in patients with fatal outcomes: a coagulation cluster (C4), a hypoxic cluster (C9), a quiescent cluster (C11).

Implications:

• Monitoring changes in the Pla-T ratio and platelet subpopulations could be valuable for predicting patient outcomes and guiding treatment decisions.
• Understanding the interplay between platelets and other immune cells may pave the way for therapies that modulate the immune response and improve patient outcomes.
• Targeting heparanase activity, which is increased in platelets during sepsis and COVID-19, could be a promising therapeutic strategy.

Bioinformatics Approaches:

• scRNA-seq Data Integration: Combined 413 samples from COVID-19, sepsis, SLE, and healthy individuals
• Focus on Platelets: Extracted platelet expression data from PBMC datasets
• Quality Control and Preprocessing: Used source data annotations, HGNChelper tool, and Harmony software
• Cell Type Identification and Clustering: Used Seurat, SingleR, and Louvain Clustering algorithm
• Differential Gene Expression Analysis: Employed MAST method from Seurat
• Machine Learning: Used XGB and DNN for biomarker identification and feature importance
• Pathway and Gene Set Enrichment Analysis: Used clusterProfiler, AddModuleScore, DOSE, and GSVA packages
• Ligand-Receptor Interaction Analysis: Based on iTALK database
• Pseudotime Trajectory Inference: Constructed using Monocle3

Key Findings:

• Lymphocyte subsets are reduced in sepsis, especially in fatal outcomes.
• Platelets and erythroid precursors emerge in late-stage fatal sepsis.
• Sepsis drives hypoxic stress associated with disease severity and dysfunctional erythropoiesis.
• Monocytes in fatal sepsis undergo immune suppression.
• Increased CD52 expression within hours of sepsis recognition is associated with improved sepsis outcomes.

Implications:

• Antiplatelet therapies may be beneficial for treating severe sepsis.
• S100A9 may be a valuable biomarker to stratify sepsis severity.
• CD52 may serve as both a biomarker for sepsis progression and a therapeutic target to promote immune homeostasis.

Bioinformatics Approaches:

• Data Processing and Quality Control: Used Cell Ranger Software Suite
• Data Integration and Batch Effect Removal: Utilized Seurat v.3
• Dimensionality Reduction and Clustering: Performed using Seurat
• Cell Type Annotation: Used a consensus-based approach combining information from canonical marker genes, SingleR, and scCATCH
• Differential Gene Expression Analysis: Employed MAST method from Seurat
• Pathway Enrichment Analysis: Conducted using clusterProfiler
• Module Score Analysis: Calculated to assess expression of predefined gene sets

Key Findings:

• AI is revolutionizing drug development, streamlining processes and increasing success rates.
• Traditional drug development faces obstacles like high costs, long timelines, and high failure rates.
• AI-driven protein structure prediction provides valuable insights for drug discovery.
• Several AI algorithms offer distinct approaches to protein structure prediction:
  • AlphaFold2 (AF2): High accuracy, relies on multiple sequence alignments
  • ESMFold: Efficient for orphan proteins, faster than AF2
  • RoseTTAFold: Unique "three-track" neural network architecture
  • OpenFold: Open-source version of AF2, offers flexibility
• Generative AI is emerging as a powerful tool in drug development.
• AI-driven drug discovery is demonstrating success beyond cancer research.

Implications:

• Widespread adoption of AI could significantly alter the pharmaceutical landscape.
• Understanding strengths and limitations of different AI algorithms is crucial for effective application.
• Integration of AI necessitates careful consideration of data privacy, security, and validation procedures.
• Continued research in AI algorithms is essential for maximizing impact in drug discovery.
• AI-driven successes across diverse disease areas underscore potential to revolutionize healthcare.

Future Directions:

• Develop AI algorithms for predicting protein dynamics and ligand-induced folding.
• Expand applications of generative AI in designing novel proteins and antibodies.
• Integrate AI-driven approaches with traditional drug discovery methods.
• Address regulatory challenges posed by AI in pharmaceutical research.
• Explore AI applications in personalized medicine and rare disease treatment.