Decoding Complexity
Decoding Complexity
This section showcases a selection of my presentations, highlighting my approach to making complex scientific concepts accessible and engaging through visual storytelling and innovative design. My goal is to foster understanding and spark curiosity in diverse audiences.
My culminating PhD work unveils a novel application of the classic Sankoff-Rousseau algorithm for the optimal inference of horizontal gene transfer (HGT) events. By re-imagining Sankoff's 'states' to represent character presence or absence, we developed a method to minimize gene losses and transfers in phylogenetic networks. This research demonstrates the limitations of standard two-state Sankoff for HGT inference and introduces an extended dynamic programming solution for optimality, alongside a heuristic that effectively identifies 'transfer highways' in real biological data, offering crucial insights into gene evolution.
Novel Algorithmic Application: Repurposes the well-known Sankoff-Rousseau algorithm for a critical problem in evolutionary biology: horizontal gene transfer inference.
Enhanced HGT Detection: Provides a theoretically optimal framework for identifying gene transfer events by addressing previous algorithmic limitations.
Biologically Meaningful Insights: The developed heuristic effectively uncovers "transfer highways" in real datasets, offering tangible understanding of gene flow and its implications for areas like antibiotic resistance.
PhD Culmination: Represents the successful completion and practical application of a theoretical model developed over three years of doctoral research.
📍Herbstseminar of Bioinformatics in Doubice, organized by the Bioinformatics Group of the University of Leipzig
Given the diverse audience of Masters and PhD students from various scientific backgrounds, and their families, my goal was to offer a high-level, informal look at the three main challenges of my PhD. I introduced the talk with a fun analogy of academic evolution as a family tree, playfully sharing my own "transfer" from Peter F. Stadler's research "subtree" to David Sankoff's. I also delved into the philosophical question of how computer science often simplifies complex problems into "yes/no" answers, akin to a game of charades, even when tackling something as intricate as evolution.
Accessible Science Communication: Successfully presented complex bioinformatics problems to a multidisciplinary audience with varied backgrounds, including non-scientists, fostering broader understanding.
New Approach: Used novel analogies (academic family tree, charades) to demystify theoretical concepts, making them relatable and engaging.
Challenging the Norm: Delivered a highly conceptual talk within the "infamous graph theory session," known for its heavy theoretical nature, demonstrating versatility in presentation style.
I had the pleasure of presenting a side project in collaboration with my former lab at CINVESTAV Irapuato. This work aims to equip biologists with a crucial tool for evaluating the quality of reconstructed evolutionary histories and reconciled gene trees, especially when a "ground truth" (like simulated data) is available. I used the metaphor of "bad vs. good" gene trees to invite the audience to rethink traditional quality metrics, like the Robinson-Fould distance and the standard Path-label distance. My presentation showed that assessing tree quality isn't just a binary choice; it exists across a spectrum depending on what information you penalize – be it just topology, or the reconciliation map itself. Our contribution uniquely leverages the entire evolutionary scenario, including topology, gene trees, and the reconciliation map, to enable a more informed assessment of reconstruction quality.
Enhanced Quality Assessment: Introduced a novel dissimilarity measure that provides a more comprehensive way for biologists to evaluate the accuracy of reconstructed gene trees against a known truth.
Beyond Simple Topology: Highlighted the limitations of traditional metrics that often only consider tree shape, emphasizing the need to also evaluate the reconciliation map for a complete picture.
Practical Tool for Biologists: Offers a valuable guide for researchers working with simulated data or when comparing different reconstruction methods in the context of evolutionary history inference.
This work delves into a variant of Perfect Transfer Networks (PTNs) with a key structural restriction: no two evolutionary cycles can share nodes, making them more "tree-like" than overly complex PTNs. My presentation focused on conveying the intuition behind two main problems: Completion (adding transfer arcs to form a galled PTN) and Compatibility (reconstructing a galled-completable tree from character data). Through numerous visual metaphors, I guided the audience to recall the set-hierarchy nature of trees and how, by sometimes splitting sets to reveal chain-like structures, we can achieve these more constrained, yet biologically insightful, galled trees.
Novel Theoretical Framework: Introduces Galled Perfect Transfer Networks as a structurally constrained, more interpretable alternative to complex PTNs for modeling evolutionary scenarios.
Intuitive Problem Unveiling: Effectively communicated complex theoretical problems (Completion and Compatibility) and definitions using visual metaphors, making abstract concepts accessible.
Rethinking Tree Structures: Highlighted the importance of viewing trees as set hierarchies and understanding how manipulating these sets can lead to simpler, galled tree structures for evolutionary analysis.
Foundation for Future Visualizations: Laid theoretical groundwork that can be leveraged to develop new, insightful visual tools for understanding evolutionary relationships.
I dove into our FPT (fixed-parameter tractable) algorithm for inferring orthology relationships. Using the classic fable of the tortoise and the hare, I illustrated how our "slow and steady" method triumphs over faster but less accurate approaches like spectral clustering. I unraveled how we adapted the NP-Hard problem of cluster editing, creatively modifying its cost function to consider "colors" – so we don't penalize edges already matching the desired cluster's attributes. Furthermore, I presented orthology clustering as a binary classification problem, drawing a clear visual link between a confusion matrix and its graph theoretical meaning, comparing edges in our inferred graph against those in a simulated "ground truth" orthology graph.
Novel Algorithmic Approach: Introduced a new FPT algorithm for orthology inference, providing a more robust and accurate method than commonly used faster alternatives.
Creative Problem Adaptation: Demonstrated an innovative adaptation of the NP-Hard cluster editing problem, tailoring its cost function to better reflect biological realities through "colorful" considerations.
Enhanced Understanding: Provided a clear, intuitive framework for understanding orthology clustering as a binary classification problem, using visual metaphors like the confusion matrix to explain graph-theoretical implications.
Improved Biological Accuracy: Highlighted how the method's performance in penalizing edges leads to more reliable orthology inferences, crucial for comparative genomics.
📍Online Seminars in Algorithms and Complexity & WABI 2022 talk from the Hasso Plattner Institute in Potsdam, Germany
This presentation focused on Perfect Transfer Networks (PTNs), a mathematical model I developed to help infer horizontal gene transfer events. My goal was to convey the core ideas and definitions of PTNs without relying on complex equations or notation. Instead, I used clear illustrations to highlight the main problems and algorithms. This work represents my first PhD paper, and it laid both the mathematical and visual groundwork for the research that followed.
Accessible Model Introduction: Successfully explained a complex mathematical model (Perfect Transfer Networks) and associated algorithms to a broad audience without resorting to intricate equations or notation.
Foundational PhD Research: Showcased the initial, foundational work of my PhD, demonstrating its role in setting the mathematical and visual basis for subsequent projects.
Clarity Through Visualization: Emphasized the power of visual communication to convey abstract algorithmic concepts and problems effectively.
Contribution to HGT Inference: Introduced a novel character-based model aimed at improving the inference of horizontal gene transfer, a crucial area in evolutionary biology.
In this talk, I explored the incredible power of visual communication in science. I broke down fundamental design concepts, illustrating each with a famous painting or artist, then drew direct parallels to concrete examples found in renowned science magazines and websites. My aim was to show how principles from the art world can be directly applied to make scientific information more impactful and accessible. I also shared simple tools and valuable resources, empowering the audience to start creating their own compelling science graphics.
Empowering Visual Communication: Demonstrated how basic design principles, inspired by famous art, can significantly enhance scientific communication.
Bridging Art and Science: Creatively linked artistic techniques to practical applications in science graphics, highlighting their shared visual language.
Practical Skill Development: Provided the audience with actionable tools and resources, enabling them to immediately begin crafting their own effective science visuals.
Promoting Accessibility: Advocated for the use of strong graphics to make complex scientific information more understandable and engaging for broader audiences.
At a recent event aimed at encouraging women in STEM, I presented "Three Goodnight Stories for Rebel Girls in Bioinformatics," a talk for the general public inspired by the popular book series. My goal was to demystify bioinformatics and offer a glimpse into the exciting research I do during my PhD. I shared a high-level overview of three of my projects, explaining the types of questions I tackle daily and how I approach solving biological problems. To make complex ideas relatable, I used the analogy of mathematical models as board games, where we set rules, and algorithms as strategies to find the best way to "win" or solve the problem.
Inspiring the Next Generation: Successfully engaged the general public, particularly aiming to encourage women and girls to consider careers in bioinformatics and STEM.
Accessible Science Communication: Translated complex PhD research into understandable, relatable "stories" using analogies like board games and strategies, avoiding jargon.
Demystifying Bioinformatics: Provided a clear, high-level overview of bioinformatics and the problem-solving approach within the field.
Role Model Representation: Leveraged the "Rebel Girls" theme to highlight the potential for women to lead and innovate in scientific research.
I gave a special talk titled "Three short stories in bioinformatics with special thanks to the women that made them possible" at Club Info, our department's official computer science talks. Bioinformatics, as a multidisciplinary science, uses computer science and mathematical tools to explore biological phenomena like evolution. In this presentation, I shared high-level insights into three of my projects focused on the evolution of gene family histories. Crucially, at the end of each "story," I took the opportunity to acknowledge and celebrate the women who were crucial for bringing these projects to life, from my inspiring teachers in genetics and mathematics to my brilliant co-authors. No prior knowledge of biology or mathematics was required; everyone curious about bioinformatics was welcome to join!
Record-Breaking Attendance: The talk achieved record attendance for Club Info, filling the room and demonstrating significant public interest in the topic.
Accessible Science Communication: Successfully introduced complex bioinformatics concepts to a general audience, making them engaging and understandable without prior specialized knowledge.
Career Inspiration: Prompted questions from students contemplating career paths in biology or computer science, indicating the talk's positive influence on career choices.
Highlighting Women in STEM: Actively celebrated and acknowledged the vital contributions of women in science and research, fostering visibility and inspiring future generations.
Interested in a collaboration or a customized presentation for your event? Feel free to contact me.