AliClu

AliClu aligns time-stamped discrete clinical sequences and clusters patients to enable stratification based on longitudinal treatment and health-state trajectories.

Key Features:

• Temporal Needleman-Wunsch (TNW): Implements a Temporal Needleman-Wunsch algorithm that aligns discrete sequences while incorporating transition time information between symbols representing therapy status, health conditions, or other states.
• Hierarchical Clustering: Applies hierarchical clustering to pairwise TNW-derived similarity scores to group patients into subgroups.
• Cluster Validation and Optimization: Uses bootstrapping resampling to assess cluster stability and to inform determination of the optimal number of clusters.
• Application to Reuma.pt EMRs: Has been applied to the Portuguese rheumatologic patient database Reuma.pt to analyze therapy switches in rheumatoid arthritis where therapies are coded as discrete symbols with durations.
• Automatic and Semi-Automatic Stratification: Supports automatic and semi-automatic stratification modes with adjustable parameters for alignment, clustering, and validation.

Scientific Applications:

• Personalized Medicine: Enables patient stratification based on temporal therapy profiles to inform personalized treatment strategies.
• Clinical Research: Facilitates identification of common trajectories and patterns within patient groups to study disease progression and treatment response.
• Longitudinal Data Analysis: Analyzes longitudinal datasets and electronic medical records (EMRs) where temporal dynamics of state transitions are critical for interpreting clinical outcomes.

Methodology:

Aligns sequences using the Temporal Needleman-Wunsch algorithm that incorporates transition times, computes pairwise TNW-derived scores, applies hierarchical clustering to those scores, and uses bootstrapping resampling for cluster validation and selection of cluster number.