ML Analysis of Urban 16S Soil and Water Data

ISEF Category: Cellular and Molecular Biology

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

A teaspoon of soil can hold thousands of microbial types, and most of them have never been grown in a dish. That makes city parks, drains, and streams a hidden map of life that changes with land use. If you sample along an urbanization gradient, you can ask whether microbial communities shift in predictable ways.

What Is It?

This project studies bacterial community DNA from soil and water samples. You collect samples from sites that differ in urban impact, send them for 16S amplicon sequencing, and then compare the results. The 16S gene acts like a barcode for bacteria, so you can group similar reads into amplicon sequence variants, or ASVs, which are high-resolution sequence types.

Think of it like sorting a giant box of mixed puzzle pieces. Sequencing gives you the pieces, and bioinformatics helps you group them into patterns. Machine learning can then look for combinations of ASVs, or environmental features, that separate low-urban and high-urban sites, or flag sequences that may represent less-studied taxa in your area.

Why This Is a Good Topic

This is a strong science fair topic because the question is measurable, local, and data-rich. You can connect microbial ecology, sequencing, and machine learning without needing to culture bacteria. The project also links to a real issue, how urbanization changes ecosystems, and you can learn how to handle sequence data, clean metadata, run classification models, and interpret community patterns.

Research Questions

• How does urbanization level affect bacterial community diversity in local soil samples? ?
• How does urbanization level affect bacterial community diversity in local water samples? ?
• What is the effect of sample type, soil versus water, on the abundance of specific 16S ASVs along the urbanization gradient? ?
• Does a machine learning model classify sites by urbanization level better than random chance using 16S features? ?
• To what extent do environmental variables such as canopy cover, impervious surface, or proximity to roads explain microbial community shifts? ?
• Which ASVs are most strongly associated with low-urban or high-urban sites? ?

Basic Materials

• Sterile sampling tubes or bags.
• Disposable gloves.
• Permanent marker and waterproof labels.
• GPS-enabled phone for site coordinates.
• Field notebook or data sheet.
• Cooler with ice packs for transport.
• Access to outsourced 16S sequencing service.
• Computer with internet access.
• Spreadsheet software for metadata tracking.
• Digital camera or phone camera for site photos.

Advanced Materials

• DNA extraction kit suitable for soil or water samples.
• PCR thermocycler.
• Agarose gel electrophoresis setup.
• Nanodrop or Qubit for DNA quantification.
• Magnetic rack if your extraction workflow uses beads.
• Bioinformatics workstation or access to a computing cluster.
• Reference database for 16S taxonomic assignment.
• Contamination controls, including extraction blanks and PCR negatives.
• Standard soil or water chemistry test kits.
• Shallow sequencing library cleanup reagents if your lab prepares libraries in house.

Software & Tools

• QIIME 2: Processes 16S reads, builds feature tables, and supports diversity and taxonomic analysis.
• R: Runs community ecology statistics, plots, and machine learning workflows.
• Python: Fits classifiers, ranks informative ASVs, and automates data cleaning.
• ImageJ: Measures site photo features, such as green cover or surface texture, if you need visual habitat variables.
• PubMed: Helps you find review articles and primary papers on urban microbial ecology and 16S analysis.

Experiment Steps

1. Define the urbanization gradient you will test, then decide how you will measure it across sites.
2. Choose paired sample types and build a metadata plan that keeps location, habitat, and collection time organized.
3. Plan your sequencing design, including technical and field controls that can catch contamination or batch effects.
4. Set up your analysis pipeline so you can convert raw reads into ASVs, taxonomy, and diversity metrics.
5. Decide which machine learning target you will predict, then choose features and a validation strategy that prevents overfitting.
6. Plan how you will compare model output with ecological interpretation, so you can link patterns back to urbanization.

Common Pitfalls

• Mixing soil and water collection methods without matching controls, which makes the data hard to compare.
• Ignoring extraction or PCR contamination, which can create fake rare taxa and distort low-abundance calls.
• Treating sequencing depth as if every sample had the same coverage, which can bias diversity comparisons.
• Feeding the model too many ASVs for too few samples, which leads to overfitting and weak validation.
• Forgetting to record site metadata carefully, which prevents you from connecting microbial patterns to the urbanization gradient.

What Makes This Competitive

A stronger project goes past a simple before-and-after comparison. You can stand out by using matched sites, clear negative controls, and a preregistered analysis plan for your model. Strong normalization, cross-validation, and an honest test set matter a lot here. The best projects also explain which taxa or community features drive the pattern, not just whether a classifier worked.

Project Variations

• Compare bacterial communities in stormwater, stream water, and nearby soil from the same urban corridor.
• Use shotgun metagenomics on a smaller sample set if you want broader taxonomic and functional signals.
• Focus on seasonal change by sampling the same sites across wet and dry conditions, then test whether season outweighs urbanization.

Learn More

• QIIME 2 Documentation: Free tutorials for 16S sequence processing, feature tables, and diversity analysis. Find it by searching the QIIME 2 official site.
• Earth Microbiome Project: Public protocols and background on large-scale microbial community sampling. Find it by searching the project name.
• NCBI SRA: Public sequence archive with example 16S datasets you can download and practice on. Find it through NCBI.
• PubMed: Search review articles on urban microbiomes, 16S amplicon analysis, and microbial ecology statistics.
• MIT OpenCourseWare Introduction to Computational Biology: Free course materials that can help you understand sequence data analysis and reproducible workflows.
• NOAA National Water Quality data resources: Background on watershed and water-quality variables that may help you design site comparisons. Find them through NOAA.