When Computers Learn Biology

How the 2018 MCBIOS Conference is Revolutionizing Science

Computational Biology Bioinformatics Big Data AI in Biology

Introduction to MCBIOS 2018: Where Data Meets Biology

Imagine training a computer to recognize the building blocks of life as effortlessly as Facebook recognizes faces in your photos. That's exactly the kind of groundbreaking work presented at the XVth Annual MidSouth Computational Biology and Bioinformatics Society (MCBIOS) conference in Starkville, Mississippi in Spring 2018. This gathering of brilliant minds centered around a powerful theme: "Genomics and Big Data."

183
Registered Participants
157
Abstracts Presented
9
Breakout Sessions

With 183 registered participants—including 97 students, 13 postdoctoral fellows, and 73 professionals—the conference served as a vibrant marketplace of ideas where biology, computer science, and statistics converged 1 . Researchers shared 157 abstracts across nine breakout sessions, tackling everything from plant genetics to precision medicine 1 .

Genomics and Big Data: Why Computers are Biology's New Microscope

What is Computational Biology?

Computational biology represents the marriage between biological data and computer algorithms to solve complex biological puzzles. Think of it this way: where traditional biologists might use microscopes to examine cells, computational biologists use algorithms to examine patterns within massive biological datasets.

Data Growth in Biology

The Big Data Challenge in Biology

The "Big Data" focus of MCBIOS 2018 highlighted a critical challenge in modern biology: we're generating biological information faster than we can interpret it. From DNA sequences to cellular images, the data deluge requires sophisticated computational approaches to extract meaningful insights.

Keynote Sessions
"Next-Gen Data Science" by Russ Wolfinger
"Real World Data and Precision Medicine" by Lawrence J. Lesko
"No-Boundary Thinking" by Steve Jennings
"Informatics Tools for Big Biologicals" by William J Welsh
"A decade of MAQC effort" by Weida Tong

1

Spotlight Experiment: Teaching Computers to See Cells

The Nucleus Segmentation Challenge

One of the most compelling stories to emerge from MCBIOS 2018 was the 2018 Data Science Bowl—a global competition to create an algorithm that could automatically identify nuclei in microscope images across different experimental conditions 7 .

Before this breakthrough, biologists had to manually adjust their analysis software for nearly every new set of images—a time-consuming process requiring specialized expertise 7 . The Data Science Bowl set out to change this by creating a universal nucleus detector that could work across various cell types, microscopes, and staining methods without human intervention.

Methodology: How Do You Teach a Computer to See?
Dataset Creation

Researchers assembled a massive training set of 37,333 manually annotated nuclei from 841 images across more than 30 different experiments 7 .

Competition Structure

The challenge was run on the Kaggle platform with 3,891 teams participating worldwide 7 .

Evaluation Method

Entries were judged on their ability to accurately segment nuclei in diverse image types, with the final evaluation containing approximately 100,000 nuclei across 3,200 images 7 .

Technical Approach

Top teams used deep-learning models—a form of artificial intelligence where neural networks learn patterns directly from data.

Remarkable Results: Computers Outperform Humans

The outcomes were staggering:

  • 85 candidate algorithms from the challenge outperformed traditional methods that required manual configuration 7
  • The top three solutions surpassed minimally-tuned classical algorithms "by a large margin" 7
  • The winning model agreed on boundary annotations more consistently with human annotators than different human annotators agreed with each other 7
Data Science Bowl Participation

Performance Comparison of Segmentation Methods

Method Type Configuration Required Accuracy Time Investment
Classical Algorithms Extensive per-experiment
Variable, often suboptimal
3-5 hours expert time
Novice User Minimal, but limited knowledge
Lowest
5 hours, poor results
Custom U-Net Models Significant (~20 hours)
Moderate
20+ hours development
Top Competition Models None
Highest
None after development

7

The Scientist's Toolkit: Key Research Reagents & Solutions

Behind every great computational discovery lies a set of powerful tools and resources. The MCBIOS conference highlighted several crucial components of the computational biologist's toolkit:

Beyond the Algorithm: Celebrating Science and Future Scientists

The MCBIOS 2018 conference wasn't just about research presentations—it reflected a vibrant, growing scientific community committed to nurturing new talent and exploring diverse biological questions.

Young Scientist Excellence

The Young Scientist Excellence Award competition highlighted the future of computational biology, with postdoctoral fellows and graduate students presenting groundbreaking work 1 .

Award Winners
  • Sundar Thangapandian, Ph.D. - Quantitative Target-specific Toxicity Prediction Model
  • Chathurani Ranathunge - Transcribed microsatellites in common sunflower
  • Brian Walker, Ph.D. - Synthesis of xanthine derivatives for PARG inhibition

1

Diverse Research Frontiers

The conference's nine breakout sessions showcased the remarkable breadth of computational biology applications 1 :

Research Areas at MCBIOS 2018

The Future of Computational Biology: No-Boundary Thinking

The proceedings of MCBIOS 2018 offer more than just a snapshot of current research—they provide a roadmap for biology's future.

Predictive Biology

Through models that can accurately simulate cellular behavior

Efficient Experiments

Through computational optimization of experimental designs 4

Accessible Analysis

Through tools that work automatically across diverse conditions

The "No-Boundary Thinking" philosophy championed at the conference points toward a future where biological discovery is limited only by our imagination, not by our analytical capabilities 1 . From optimizing associative learning experiments through computational methods 4 to creating personalized heart cell models for drug testing 9 , the work presented at MCBIOS 2018 demonstrates that the most exciting breakthroughs happen at the intersections between fields.

The Future is Interdisciplinary

As we look ahead, the type of research showcased at MCBIOS 2018 promises to accelerate our understanding of life's fundamental processes while delivering practical benefits in medicine, agriculture, and environmental science.

References