Master’s thesis topics at the intersection of pharmacy, computer science, and data science available

Are you studying medical informatics, computer science, data science, pharmacy, or a related field and looking for a clinically relevant topic for your master’s thesis? The INTERPOLAR project of the Medical Informatics Initiative is now offering five master’s theses that link research and clinical application.

The topics are situated in basic research and methodological development in digital medication safety and artificial intelligence (AI). Supervision is possible across pharmacy, data science, and computer science. You will work with current approaches in AI, causal modeling, and knowledge representation while gaining insights into clinical processes. This allows you to gain project experience, build connections, and develop key competencies for careers in digital health already within the scope of your thesis.

These are the topics:

Topic 1:

Title: Development of a Semantic Pipeline for Structured Knowledge Representation of Drug Information

Background:

Drug information is often available in textual form (e.g., professional information, guidelines, and study reports). To derive justifiable and verifiable decision logic from this information, the content must be formalized, semantically enriched, and made machine-readable.

Objectives:

• Establish a pipeline for extracting clinically relevant concepts (e.g., dosage, interactions, contraindications)
• Preprocessing, validating, and refining annotations
• Concept and rule extraction (e.g., dosages, contraindications, and drug interactions)
• Semantic representation in knowledge models (e.g., FHIR MedicationKnowledge, OWL, JSON-LD)
• Integration of reasoning logic (“reasoning as a model”) to derive clinical rules
• Evaluation based on case studies
• Mapping in FHIR knowledge artifacts, Contextual Query Language (CQL), and ontologies

Requirements:

• Interest in natural language processing (NLP), knowledge representation, and semantic technologies
• Basic knowledge of Python, Resource Description Framework (RDF), or FHIR
• Willingness to work interdisciplinarily with clinical data

Topic 2:

Title: Modeling and Analyzing Clinical Decision-Making Reasoning with Reasoning Graphs

Background:

Clinical decisions, especially those related to medication management, are based on explicit and implicit reasoning. These can be modeled formally as reasoning graphs to identify patterns, fallacies, and strategies in decision-making processes.

Objectives:

• Collection of pharmacist-based case analyses
• Structured recording of reasoning to identify medication problems and assess their relevance
• Modeling reasoning graphs (nodes: observation, hypothesis, and action; edges: supports, contradicts, and leads to), annotating clinical or pharmaceutical case reasoning where applicable
• Development of a schema for “reasoning as a model” (nodes: observation, hypothesis, action)
• Consolidation of individual graphs into a generalized reasoning and knowledge graph
• Quantitative and qualitative analysis of decision-making logic
• Comparison of human vs. machine reasoning patterns (e.g., Large Language Models, knowledge graphs)

Requirements:

• Interest in cognitive modeling, graph reasoning, and explainable AI
• Basic knowledge of Python, Neo4j, Graphviz, or comparable tools is helpful
• Ability to think analytically and interdisciplinarily

Topic 3:

Title: The Extraction, Formalization, and Justifiability of Pharmaceutical Rules for Medication Decisions

Background:

Clinical pharmaceutical recommendations (e.g., dosage adjustments and interaction warnings) are based on rules that are rarely explicitly justified or validated. A machine-readable, explainable rule base is necessary to make clinical decision support comprehensible.

Objectives:

• Systematic extraction of pharmaceutical rules from guidelines and literature (e.g., contraindications). Contraindications are circumstances that prohibit a medical measure because it could lead to health damage or side effects, for example.
• Classification and mapping to standardized terminologies (e.g., SNOMED CT, ATC, and ICD-10)
• Formal mapping as a rule graph or in CQL/FHIR PlanDefinition
• Integration of reasoning information (reasoning components)
• Development and testing of a prototype for the algorithmic mapping of contraindication rules (e.g., evidence cards that link data, rules, and reasoning)
• Evaluation of the explainability and consistency of these rules using clinical scenarios (validation by subject matter experts)

Requirements:

• Degree in pharmacy, data science, or computer science
• Interest in rule modeling, semantic interoperability, and evidence representation
• Basic knowledge of Python, text mining, and/or knowledge modeling
• Enjoyment of interdisciplinary collaboration between computer science, pharmacy, and clinical practice

Topic 4:

Title: Methodological Approaches to Integrating Pharmacological Modeling and Observational Data

Background:

Pharmacological models (e.g., PopPK/PD models) and routine data reflect different sources of evidence. Integrating these models requires a methodological approach that harmonizes their structures, uncertainties, and scales.

Objectives:

• Comparison of methodological strategies for integrating structured and unstructured data sources
• Development of a framework for consistency checking of pharmacological and observational evidence
• Evaluation of sources of bias and uncertainties
• Application to exemplary drug groups (e.g., anticoagulants and immunosuppressants).

Requirements:

• Degree in data science, biostatistics, computational pharmacology, or computer science
• Knowledge of statistics, Bayesian modeling, or machine learning
• Interest in evidence integration and model validation

Topic 5:

Title: Development of Data-Driven Simulation Models (Digital Twins) for Medication Safety

Background:

Digital twins can simulate complex drug effects and interactions. Integrating genetic, clinical, and pharmacological data creates a framework for model-based hypothesis generation and risk assessment.

Objectives:

• Design and implementation of a generic simulation framework
• Integration of patient-specific variables (e.g., genetic markers, organ function, and co-medication)
• Development of models to predict dose-response relationships and the risk of adverse drug interactions
• Methodological evaluation of model quality, generalizability, and explainability

Requirements:

• Degree in computational biology, data science, or medical informatics
• Interest in simulation, AI, pharmacometrics, or model validation
• Experience with Python, R, or simulation frameworks (e.g., PyMC, Stan)

Advantages for Students:

The topics strengthen methodological skills at the interface of explainability, evidence integration, and model development. They are theoretically relevant to knowledge representation and causal modeling while being practically important for medication safety and explainable AI. They also combine natural language processing, knowledge engineering, and FHIR—key competencies for careers in health data science, clinical AI, and the digital health sector.

Are you interested in one of the topics? Contact us with your topic request at info@smith.care!