ISEF Project Ideas by Category (2027): 60 Research Questions That Can Actually Win

An ISEF project idea is not a topic. "Microplastics" is a topic. "Does polyester microfiber concentration in household laundry effluent change with wash temperature, and can a mesh filter cut it by more than half?" is a project. Every idea below is written as a research question with a method attached, because the gap between the two is where most students lose a year.

The ideas are grouped by ISEF category. Under each category is a note on how accessible it is to a student without a university lab, because that is the constraint that decides what you can actually build. If you are not sure whether an idea passes the judges' bar, run it through What Makes a Project ISEF-Worthy before you start.

Cellular and molecular biology

Access: hard without a lab. Most projects need cell culture, PCR or microscopy. The computational versions are fully accessible.

1. Does gene expression in publicly available single-cell RNA-seq datasets identify a marker that separates early from late-stage tumor samples? (Computational, uses GEO data.)
2. How does caffeine concentration affect the rate of cell division in yeast, measured by optical density over 48 hours? (Wet lab, school-feasible.)
3. Can a CRISPR guide RNA design tool be improved by adding a secondary structure penalty, validated against published efficiency data? (Computational.)
4. Does the mitochondrial DNA copy number in published datasets correlate with reported patient age across tissue types? (Computational.)

Biomedical and health sciences

Access: moderate. Public health data is abundant; clinical data needs a mentor with access.

5. Which features in a public ICU dataset best predict readmission within 30 days, and does a simple model match a complex one?
6. Does the timing of insulin dosing in a public type 1 diabetes CGM dataset predict overnight hypoglycemia?
7. Is there a measurable relationship between local air quality index and asthma emergency visits in open city health data?
8. Can smartphone accelerometer data classify gait changes associated with fatigue in healthy volunteers?

Computational biology and bioinformatics

Access: fully accessible. This is the most open ISEF category for a student with a laptop and a mentor.

9. Can a machine learning model trained on protein sequence alone predict thermostability, benchmarked against a published method?
10. Which microbiome features in a public gut sequencing dataset separate inflammatory bowel disease patients from controls, and are the results stable across cohorts?
11. Does a graph neural network outperform a random forest at predicting drug-target interactions on a public benchmark?
12. Can k-mer frequency profiles classify viral sequences by host species without alignment?

Chemistry

Access: moderate. Many projects are feasible with school equipment; instrumental analysis needs a lab.

13. Which plant-derived dye produces the most stable color on cotton after repeated washing, and what mordant concentration maximizes it?
14. How does the ratio of citric acid to sodium bicarbonate affect the reaction rate in a home-built calorimeter, and does the data fit a known rate law?
15. Can a low-cost colorimetric sensor built from a smartphone camera quantify nitrate in water samples against a laboratory standard?
16. Does the addition of a specific polymer improve the shelf life of a biodegradable bioplastic made from starch, measured by tensile strength over time?

Earth and environmental sciences

Access: fully accessible. Field data you collect yourself is a strength, not a weakness.

17. How does tree canopy cover, measured from satellite imagery, relate to surface temperature across neighborhoods in your city?
18. Does microplastic concentration in a local river change with distance from a wastewater treatment outfall?
19. Can soil moisture sensors on a low-cost network predict irrigation need more accurately than a fixed schedule in a school garden?
20. What is the relationship between rainfall intensity and turbidity in a local stream, and does it differ between urban and rural reaches?

Embedded systems

Access: accessible with an Arduino or Raspberry Pi budget.

21. Can a low-power embedded sensor node detect early signs of a leak in household plumbing from acoustic signatures?
22. Does a custom wake-word detector running on a microcontroller achieve usable accuracy below a fixed power budget?
23. Can an embedded camera system count pollinators visiting a plant accurately enough to replace manual observation?

Energy: sustainable materials and design

Access: moderate. Small-scale builds are feasible.

24. How does the angle and spacing of a bifacial solar array affect total daily output on a rooftop, measured over a season?
25. Can a phase-change material in a wall panel reduce indoor temperature swing in a scale-model room?
26. Does adding a specific nanoparticle to a dye-sensitized solar cell improve efficiency, measured with a home-built current-voltage rig?

Engineering technology: statics and dynamics

Access: accessible with a 3D printer and basic test rig.

27. Which lattice geometry in a 3D-printed structure gives the highest strength-to-weight ratio under compression?
28. Can a passive damping design reduce vibration in a drone frame, measured by accelerometer data in flight?
29. How does bridge truss design change under asymmetric loads, tested with a scale model and strain gauges?

Environmental engineering

Access: accessible. Real local stakeholders make these projects stand out.

30. Does a biochar filter remove heavy metals from runoff more effectively than activated carbon at the same cost?
31. Can a constructed wetland model reduce nitrogen in simulated agricultural runoff, and what plant species works best?
32. Which low-cost method best reduces fine particulate matter indoors in a classroom, measured with a calibrated sensor?

Materials science

Access: moderate.

33. How does the concentration of a natural additive affect the biodegradation rate of a PLA film in soil?
34. Can a self-healing concrete mix using bacterial spores recover compressive strength after cracking at small scale?
35. Does electrospinning parameter choice change the fiber diameter of a biopolymer scaffold, measured with microscopy?

Mathematics

Access: fully accessible. Mentor quality matters more here than anywhere.

36. What is the distribution of a specific graph invariant across random graphs of a given class, and can a conjecture be proved for small cases?
37. Can an improved bound be found for a known combinatorial problem using a computational search guided by a theoretical argument?
38. How does a new voting rule behave under strategic manipulation, analyzed with simulation and proof?

Microbiology

Access: hard without a lab, though some projects are feasible with sterile technique at school.

39. Which common household antimicrobial produces the largest zone of inhibition against a non-pathogenic bacterium, and does resistance develop over serial passages?
40. Does soil microbial diversity, measured by colony morphology counts, differ between organic and conventional farm plots?
41. Can bacteriophages isolated from local water samples lyse a laboratory strain, and how specific are they?

Physics and astronomy

Access: moderate. Public astronomical data makes astronomy projects fully accessible.

42. Can transit timing variations in public TESS light curves reveal an additional planet in a known system?
43. How does the drag coefficient of a sphere change with surface texture, measured in a home-built wind tunnel?
44. Does a machine learning model classify variable star types from light curves as accurately as the published pipeline?
45. What is the relationship between string tension and inharmonicity in a guitar string, measured with spectral analysis?

Plant sciences

Access: accessible. Long timelines, so start early.

46. Does inoculating seedlings with a mycorrhizal fungus change drought tolerance, measured by biomass under controlled water stress?
47. How does light spectrum affect the production of a specific pigment in a leafy green grown under LEDs?
48. Can a hyperspectral or multispectral camera detect nutrient deficiency before it is visible to the eye?

Robotics and intelligent machines

Access: accessible with a budget. A built system is inherently original.

49. Can a low-cost robotic arm learn a pick-and-place task from demonstration faster with a specific reward design?
50. Does a swarm of simple robots complete a coverage task more efficiently with a specific communication rule?
51. Can a soft robotic gripper made from a specific material handle delicate produce with less damage than a rigid one?

Systems software

Access: fully accessible.

52. Does a modified cache replacement policy reduce miss rates on real workload traces compared with LRU?
53. Can a static analysis tool detect a specific class of concurrency bug in open-source projects with low false positives?
54. How does a custom scheduler affect latency for mixed workloads on a Raspberry Pi cluster?

Behavioral and social sciences

Access: fully accessible. Judges reward rigorous statistics.

55. Does the framing of a climate message as local versus global change stated willingness to act, in a preregistered survey experiment?
56. How does social media use pattern, measured from exported screen-time data, relate to self-reported sleep quality in teenagers?
57. Does a brief mindfulness intervention change performance on an attention task in a randomized classroom study?

Machine learning and artificial intelligence

Access: fully accessible. Baselines and benchmarks make execution easy to prove.

58. Can a small language model fine-tuned on a specific domain match a large general model on a narrow task at a fraction of the cost?
59. Does a specific data augmentation strategy improve robustness of an image classifier to distribution shift, measured on a standard benchmark?
60. Can an interpretable model match a black-box model on a tabular medical prediction task, and what does it reveal?

Which categories are most open to students without a lab

If you do not have lab access, the categories where ISEF-worthy projects are most feasible are computational biology, environmental sciences, environmental engineering, systems software, machine learning, mathematics, behavioral sciences, and astronomy using public data. Cellular biology, microbiology and chemistry projects can win, but almost always with a mentor who provides the lab.

Every idea above gets better with a mentor who knows the literature, because the difference between a good question and a winning one is knowing what has already been answered. That is the first thing a YRI Fellowship mentor does with a student, before any data is collected.

Frequently Asked Questions

What are the ISEF categories? There are 21: animal sciences, behavioral and social sciences, biochemistry, biomedical and health sciences, biomedical engineering, cellular and molecular biology, chemistry, computational biology and bioinformatics, earth and environmental sciences, embedded systems, energy, engineering technology, environmental engineering, materials science, mathematics, microbiology, physics and astronomy, plant sciences, robotics and intelligent machines, systems software, and translational medical science.

What are good ISEF cellular biology project ideas? Computational projects using public single-cell or bulk RNA-seq data are the most accessible, for example identifying expression markers that separate disease stages. Wet lab projects such as measuring cell division rates in yeast under different conditions are feasible at school with careful technique.

How do I choose an ISEF project idea? Pick a category where you have or can get the tools, find a specific gap by reading three recent papers, frame the question so it has a numerical answer, and check that the method appears in the literature. How to Choose a Research Topic walks through the process.

Can a computer science project win ISEF? Yes. Systems software, machine learning, robotics and computational biology produce ISEF finalists every year, and they are the categories most accessible without a lab.

How long does an ISEF project take? Plan on six to nine months from question to fair, including time to collect more data than you think you need and to write the paper before the poster. Starting in the summer before the fair season is the norm for finalists.