Olfactory Mixtures Predictions Evaluation

Validation and scoring scripts for the Olfactory Mixtures Predictions DREAM Challenge and DREAM Olfactory Mixtures Prediction Challenge 2025. For collecting writeups, see writeup-workflow.

Evaluation Overview

The challenge is split into two phases:

• Leaderboard phase: participants submit a prediction file that will be evaluated against a validation dataset.

• Final phase: participants submit a prediction file that will be evaluated against a test dataset.

Metrics returned and used for ranking are:

2024

• Mean Root Mean Square Error (mRMSE): This metric measures the average difference between predicted and actual values across all data points. A lower mRMSE indicates a model with higher accuracy, as it demonstrates a smaller average error between the model's predictions and the actual outcomes.

• Pearson Correlation: This metric assesses the linear relationship between the predicted and actual values. A higher Pearson correlation value signifies a strong positive relationship, indicating that the model's predictions align closely with the actual data.

2025

• Cosine distance: This metric measures the difference in orientation between predicted and actual vectors, independent of their magnitudes. A lower cosine distance indicates greater similarity in direction, demonstrating that the model effectively captures the overall pattern or profile of the data.

• Pearson correlation: This metric evaluates the strength of the linear relationship between predicted and actual values. A higher Pearson correlation indicates a stronger positive linear relationship, suggesting that the model's predictions closely match the observed data.

By combining these two metrics for each year, the challenges provide a comprehensive evaluation of each model's accuracy and predictive power, ensuring that both the magnitude of the prediction errors and the consistency of the predicted trends are taken into account.

Usage - Python

Validate

python validate.py \
  -p PATH/TO/PREDICTIONS_FILE.CSV \
  -g PATH/TO/GOLDSTANDARD_FILE.CSV [-o RESULTS_FILE]

If -o/--output is not provided, then results will print to STDOUT, e.g.

{"submission_status": "VALIDATED", "submission_errors": ""}

What it will check for:

• Four columns named Dataset, Mixture_1, Mixture_2, and Predicted_Experimental_Values (extraneous columns will be ignored)
• Dataset values are strings
• Mixture_1 and Mixture_2 are integers
• disease_probability values are floats between 0.0 and 1.0, and cannot be null/None
• There is exactly one prediction per mixture (so: no missing or duplicated combination of Dataset + Mixture_1 + Mixture_2)
• There are no extra predictions (so: no unknown combination of Dataset + Mixture_1 + Mixture_2)

Score

python score.py \
  -p PATH/TO/PREDICTIONS_FILE.CSV \
  -g PATH/TO/GOLDSTANDARD_FILE.CSV [-o RESULTS_FILE]

If -o/--output is not provided, then results will output to results.json.