MedGeroX Tutorials
Learn how MedGeroX uses Retrieval-Augmented Generation, scientific embeddings and semantic retrieval to explore evidence in ageing, geroscience and precision geromedicine.
Getting Started
MedGeroX allows you to ask natural-language questions about ageing and geromedicine and receive responses based on the scientific knowledge available to the system.
You do not need to know the terminology used in the original research documents. The retrieval system searches for semantically relevant information based on the meaning of your question.
How to Ask a Good Question
The quality of the retrieved context depends partly on how clearly the question defines the scientific topic.
Understanding the RAG Pipeline
Retrieval-Augmented Generation, or RAG, is the core architecture behind MedGeroX.
Step-by-Step: How MedGeroX Works
Scientific documents are collected
Research papers and other scientific material relevant to ageing and geromedicine form the knowledge base used by the retrieval system.
Documents are split into chunks
Large documents are divided into smaller sections so that the system can retrieve the passages most relevant to a particular question.
Text is converted into embeddings
Each chunk is converted into a numerical representation called an embedding. Similar scientific concepts tend to have similar positions in the embedding space.
Embeddings are indexed
The embeddings are stored in a vector database, allowing the system to perform semantic similarity searches.
The user's question is embedded
When you submit a question, the same embedding process is applied to the query.
Relevant evidence is retrieved
The query vector is compared with the vectors stored in the knowledge base. The most semantically relevant passages are retrieved.
The language model generates the response
The retrieved passages are supplied as context to the language model, which synthesizes the information into a natural-language answer.
What Are Embeddings?
An embedding converts text into a numerical vector that captures aspects of its semantic meaning.
For example, the concepts
cellular senescence,
senescent cells and
age-related senescence
can be represented as vectors whose positions reflect
their semantic relationships.
This allows MedGeroX to search by meaning rather than depending only on exact keyword matches.
How Semantic Retrieval Works
Traditional keyword search looks for matching words. Vector search instead compares numerical representations of the query and document chunks.
User Query
↓
Query Embedding
↓
Vector Similarity Search
↓
Top Relevant Chunks
↓
Context for LLM
This is particularly useful for scientific questions because the terminology used in a question may differ from the wording used in the underlying research literature.
What Does LangChain Do?
LangChain provides the orchestration layer connecting the retrieval and generation components of the application.
Instead of manually managing every step between the user question, vector search, retrieved documents and language model, LangChain can coordinate these components as a retrieval-generation pipeline.
User Query
↓
Retriever
↓
Relevant Documents
↓
Context
↓
LLM
↓
Response
Example: Exploring Cellular Senescence
Suppose a user asks:
MedGeroX converts the question into an embedding and searches the scientific knowledge base for relevant passages discussing cellular senescence, ageing mechanisms and related biological processes.
The retrieved information is then supplied to the language model as contextual evidence. The model uses that context to formulate the final response.
Suggested Research Topics
- Hallmarks and mechanisms of ageing
- Cellular senescence
- Biological-age biomarkers
- Epigenetic ageing
- Inflammaging
- Mitochondrial dysfunction and ageing
- Age-related multimorbidity
- Healthy longevity
- Geroscience interventions
- Precision geromedicine
Responsible Use
MedGeroX is designed for scientific exploration and educational use. Generated responses depend on the quality, coverage and currency of the underlying knowledge base.