Genome Assembly and Annotation

Mentors: Student: Dineshkumar Anbalagan

The goal of this project is to provide a Model Context Protocol (MCP) server that is suitable for production and exposes expected REST endpoints from the European Nucleotide Archive (ENA) as organised, schema-driven tools for AI systems. Although ENA offers robust REST APIs for biological data access, direct integration with AI agents frequently results in inconsistent queries, low reproducibility, and lack of validation. In order to overcome this, the project implements an MCP-based interface layer that ensures predictable execution, provides a strict input/output schemas, and enables secure, secure communication with ENA services. A dynamic tool registry for flexibility, Pydantic for schema validation, httpx for asynchronous API communication, and a response normalisation layer for consistent outputs are all features of the FastMCP-built system. To ensure scalability and production readiness, it will also include a comprehensive error handling, testing, and containerised deployment using Docker and Kubernetes. A fully functional MCP server with several ENA capabilities (such as study, sample, and run queries), thorough test coverage, developer-friendly documentation, and a deployable system that is simple to interface with AI clients are some of the key deliverables. This project will enable more dependable AI-driven scientific workflows by improving the reproducibility, traceability, and accessibility of biological data.

Mentors: Student: Szabolcs Vidám

My proposal is to develop a scalable computational pipeline to construct, annotate, and visualise Sequence Similarity Networks (SSNs) for a representative subset of approximately 10 million MGnify proteins. The pipeline will take FASTA files as input and utilise MMseqs2 to efficiently compute all-against-all pairwise sequence similarities. It will then use the Python library NetworkX to construct the mathematical graphs, integrating biological metadata, such as the biome-of-origin, directly into the node attributes to enable contextual environmental exploration. To ensure scalability and avoid memory bottlenecks, the pipeline will leverage NetworkX's backend dispatching architecture to route computationally heavy operations to high-performance GPU libraries like cuGraph. Finally, the graph data will be exported into simple tabular formats compatible with large-scale visualisation platforms like Cytoscape and Cosmograph. The deliverables of this project will include the core Python/Bash SSN generation pipeline, a MGnify metadata annotator, a graph exporter utility, a fully rendered 10-million node reference dataset, and comprehensive documentation alongside an integration test suite.

Mentors: Student: Soomin Lee

The Ensembl Plants and Metazoa platforms face a significant metadata gap where critical biological context, such as ploidy, strain, and sex, is often documented in peer-reviewed literature but missing from formal INSDC sequence archives. This lack of structured metadata, particularly ploidy, creates technical bottlenecks in downstream comparative genomics and requires labor-intensive manual curation. To address this, I propose developing a standalone, Python-based "Retrieve-Extract-Verify" module that leverages Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) to automate metadata extraction from scientific papers. This tool will utilize literature APIs like Europe PMC, employing section-aware parsing and confidence scoring to ensure high data integrity with minimal human intervention. My previous experience in building Transformer-based bioprocess optimization models and genome-level embeddings directly informs my approach to handling complex biological datasets. The final outcome will be a ready-to-deploy Nextflow module designed for seamless integration into Ensembl’s production pipeline and provide the global research community with richer, more reliable genomic metadata.

Mentors: Student: Esra Ersan

Perturbation biology datasets across CRISPR screens, MAVE, and scPerturb-seq remain siloed in incompatible formats, making cross-modal reasoning about genetic perturbations nearly impossible at scale. This project builds a perturbation-aware LLM by fine-tuning BioMedLM on a curated multimodal training corpus derived from the EMBL-EBI Perturbation Catalogue, enabling natural language queries such as "what happens if gene X is knocked out in cell type Y?" Deliverables include a multimodal training corpus, a fine-tuned LLM prototype, a reproducible evaluation pipeline with gene-level splits distinguishing genuine generalisation from memorisation, and full open-source documentation for reuse by the Perturbation Catalogue team.

Mentors: Student: UninterpretableEvolvingBlackBox

Ensembl VEP's web interface offers dozens of configuration options for variant annotation, overwhelming new users and generating recurring helpdesk queries. Ask VEPai is a locally-hosted, open-source AI assistant that translates natural-language descriptions into recommended VEP configurations with justifications. The system uses Retrieval-Augmented Generation (RAG) grounded in a curated knowledge base, with a deterministic constraint checker that catches species violations and option conflicts the model misses. Beyond the project brief's scope of option labelling, training data, and a prototype model, this proposal adds four components: (1) the constraint checker, empirically validated across three model sizes, (2) a quantitative evaluation framework with leave-one-out methodology and multi-run statistical replication, (3) a structured JSON schema mapping recommendations directly to VEP web form sections for "click to apply" integration, and (4) an interpretability layer where every recommendation includes a source citation traceable to the knowledge base. A working demo validated across Qwen 2.5 3B/7B/14B shows +19–30% Enable F1 from the knowledge base, with all code and results published. Core deliverables: expanded KB (~55 options), 15–20 gold-standard examples, RAG pipeline, constraint checker, JSON schema, evaluation report, and documentation. My availability (~400 hours) supports extending to a 350-hour Large project, adding QLoRA fine-tuning, attribution testing, FastAPI wrapper, and VEP output explainer.