ML-Based Taxon Classification for MalariaGEN Data

[Aadik1ng]

Early Contribution and GSoC Interest: ML-Based Taxon Classification for MalariaGEN Data

---

Issue

Hello maintainers,

I have recently submitted a pull request (linked below) and would like to express my interest in contributing further to the project, particularly in the context of Google Summer of Code.

PR: #1164

---

Summary of Current Contribution

The submitted PR introduces utility methods to improve taxon and species discovery within sample metadata:

• list_taxons()
• list_species_calls()

These methods provide a simple, read-only interface for extracting unique taxonomic and species-level information, reducing the need for manual aggregation and improving usability during exploratory analysis.

The implementation includes:

• Robust handling of missing columns and varying data formats
• Numpydoc-compliant documentation with usage examples
• Comprehensive test coverage across multiple datasets (ag3, af1, adir1, amin1)
• 20 additional test cases, all passing successfully

---

Motivation

This contribution is intended as an initial step toward improving data discovery workflows within the library. It addresses a common friction point when working with sample metadata and aims to make the API more intuitive and efficient for users.

---

Interest in GSoC and Proposed Direction

I am interested in contributing to MalariaGEN through Google Summer of Code and have been developing a project proposal aligned with the ecosystem.

Proposed Project:

Machine Learning-Based Taxon Classifier for Genomic Data

The core idea is to build a lightweight classification system that can identify major Anopheles taxonomic groups directly from raw FASTQ sequencing reads.

Key Objectives:

• Develop a feature extraction pipeline from raw sequencing data (e.g., k-mer based representations)
• Train robust classification models (e.g., Random Forest, Gradient Boosting, or neural networks)
• Avoid dependency on variant-calling pipelines to reduce computational overhead
• Provide a scalable inference system via API and cloud integration

Expected Impact:

• Simplifies taxonomic classification workflows
• Reduces computational requirements
• Improves accessibility for users working with raw sequencing data
• Complements existing tools in the ecosystem by enabling earlier-stage classification

---

Request for Feedback

I would greatly appreciate feedback on:

• The submitted pull request and its design decisions
• Whether the proposed project direction aligns with current priorities
• Suggestions for additional areas where I can contribute effectively

Thank you for your time and consideration.