classification metric · BEGINNER LESSON
AUPRC
AUPRC means area under the precision-recall curve. It is a single number from 0 to 1 that summarizes how well a classifier finds the positive class while avoiding false positives, especially when positive cases are rare.
What it is and what is measured
AUPRC is a summary of the precision-recall curve, often called a PRC. A classifier is a rule or model that assigns scores or labels to items, such as cells, genes, or patients, and the PRC compares those predictions with the true classes. Precision measures the fraction of predicted positives that are truly positive. Recall measures the fraction of truly positive items that are found. AUPRC measures the area under the curve made by plotting precision against recall. A larger AUPRC means better classification performance. This metric is especially useful when the positive class is small compared with the negative class, because it focuses on performance for the rare target class rather than being diluted by many negatives. [0,2]
[1] Commonly used software tools produce conflicting and overly-optimistic AUPRC values [3] Commonly used software tools produce conflicting and overly ... - NIH
How it is built
To make a PRC, the model’s scores are sorted and each distinct score can be used as a threshold. A threshold is a cutoff: items at or above it are called positive, and the rest are called negative. At each cutoff, precision and recall are computed, producing anchor points in precision-recall space. Those points are connected to form the curve. AUPRC then summarizes the curve’s shape with an area calculation. Different software tools may connect points in different ways, which can change the final AUPRC and even change how classifiers are ranked. [0,2]
[1] Commonly used software tools produce conflicting and overly-optimistic AUPRC values [3] Commonly used software tools produce conflicting and overly ... - NIH
Why it is used
AUPRC is often used when classes are imbalanced, meaning one class has far fewer examples than the other. In that setting, AUROC, the area under the receiver operating characteristic curve, can look good even when early positive predictions are weak. The sources say AUPRC is more sensitive than AUROC to errors among the top predictions, so it can better expose problems when the task depends on finding rare positives, such as rare cell types, cancer genes, or differential expression genes. [0,2,1]
[1] Commonly used software tools produce conflicting and overly-optimistic AUPRC values [2] AUPRC: a metric for evaluating the performance of [3] Commonly used software tools produce conflicting and overly ... - NIH
Limitations and caveats
AUPRC depends on how the curve is constructed, so software choices matter. The sources report that some tools use methods that can be overly optimistic, especially with ties in scores, and that different tools can give different AUPRC values for the same data. AUPRC also does not remove the need to understand the underlying task: a high value can still hide weaknesses outside the positive-class search. In differential gene expression studies, for example, AUPRC is useful because it directly measures precision and recall for DEGs, but it should be interpreted in the context of the dataset and the evaluation method. [0,1]
[1] Commonly used software tools produce conflicting and overly-optimistic AUPRC values [2] AUPRC: a metric for evaluating the performance of
SEE THE IDEA
How AUPRC is formed
The model gives each item a score that ranks how positive it seems.
Each cutoff turns scores into predicted positives or negatives.
Each cutoff gives one point: how many predicted positives are correct, and how many true positives were found.
Connect the points and summarize the curve with the area under it.
Each step uses the previous one: scores become thresholded predictions, predictions become precision/recall points, and the curve becomes AUPRC.
AUPRC is the area under a precision-recall curve built from model scores at multiple thresholds.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.