non convex sampling

Problem

Flux sampling is challenging, and even when uniformity and strict statistical criteria are achieved [1], other factors can lead to biologically non-meaningful samples. For example, loopy flux distributions may continuously consume energy, resulting in thermodynamically infeasible solutions [2].

In this project the student has to exploit whether such aspects can be limited by integrating omics data in a metabolic model.

If this is the case, then the student will develop further dingo approaches for leveraging sampling for biological inference. Otherwise, the student will work towards assessing the scalability and integration potential of non-convex sampling approaches (e.g., [3]) within the dingo library.

Sampling points in non-convex domains is a very challenging problem and we might need to introduce several ad hoc tools to make it efficient.

References

[1] Chalkis, Fisikopoulos, Tsigaridas, Zafeiropoulos - dingo: a Python package for metabolic flux sampling

[2] De Martino - Scales and multimodal flux distributions in stationary metabolic network models via thermodynamics

[3] Saa, Zapararte, Drovandi, Nielsen - LooplessFluxSampler: an efficient toolbox for sampling the loopless flux solution space of metabolic models

Difficulty: Hard

Size

Large (350 hours)

Skills

• Required: Python, data integration, basics on microbial metabolism
• Preferred: linear algebra, optimization theory

Expected impact

This project aims to establish the dingo Python library as a robust tool for real-world flux sampling studies, offering users a practical guide to tackling fundamental biological challenges commonly encountered in such analyses.

It will also make clear whether non-convex sampling approaches are required in the case of metabolic models. If yes, novel sampling methods will be introduced as part of the dingo library.

Mentors

• Haris Zafeiropoulos <haris.zafr at gmail.com> is post-doctoral researcher at KU Leuven in the Lab of Microbial Systems Biology, focusing on microbial interactions through data integration and reverse ecology approaches. He has participated in GSoC as both a student and a mentor with the GeomScale group, contributing to the dingo library (2021–2025).

• Vissarion Fisikopoulos <vissarion.fisikopoulos at gmail.com> is an international expert in mathematical software, computational geometry, and optimization, and has previous GSOC mentoring experience with Boost C++ libraries (2016-2020) and the R-project (2017-2019).

• Apostolos Chalkis <tolis.chal at gmail.com> is a Research Engineer at Quantagonia GmbH. He is an expert in statistical software, computational geometry, and optimization, and has previous GSoC student experience (2018 & 2019) and mentoring experience with GeomScale (from 2020 to 2023).

Tests

• Easy: Download, compile and run a flux sampling analysis on the core model of E.coli (e_coli_core) requiring: i. for optimal biomass growth, ii. for at least half of the optimal, and iii. setting biomass free.
• Medium: Exploit dingo's functionalities to come up with reaction clusters differentiating in each case.
• Hard: Try to interpret your findings. Then, describe how you would set a pipeline to address the problem described above.