likelihood
Definition
Likelihood is a statistical measure quantifying the probability of observing specific data given a particular model or hypothesis. In biological research, likelihood functions assess how well experimental observations fit proposed mechanisms, such as evolutionary models, gene regulatory relationships, or protein interaction networks. Unlike probability, which describes the chance of future events, likelihood evaluates how plausible a model is given existing data. Maximum likelihood estimation (MLE) identifies parameter values that make observed data most probable, making it fundamental for phylogenetic analysis, sequence alignment, parameter estimation in systems biology models, and inferring causal relationships in biological networks. Higher likelihood values indicate better model-data agreement.
Visualize likelihood in Nodes Bio
In Nodes Bio, likelihood scores can weight network edges to represent confidence in biological relationships. Researchers can visualize protein-protein interactions or gene regulatory connections with edge thickness or color intensity proportional to likelihood values from experimental data. This enables filtering low-confidence connections, identifying high-probability pathways, and comparing alternative network topologies based on their likelihood of explaining observed phenotypes or expression patterns.
Visualization Ideas:
- Weighted protein-protein interaction networks with edge thickness representing likelihood scores
- Gene regulatory networks colored by maximum likelihood estimates of transcription factor binding
- Phylogenetic trees with branch support values derived from likelihood ratio tests
Example Use Case
A cancer genomics team investigates signaling pathways driving tumor growth. They measure phosphorylation states of 50 proteins across 100 patient samples and use maximum likelihood estimation to infer kinase-substrate relationships. Each potential regulatory connection receives a likelihood score based on correlation patterns and known biochemical constraints. By visualizing only edges exceeding a likelihood threshold, they identify a previously unknown feedback loop involving MAPK and PI3K pathways that explains drug resistance patterns in 30% of patients, guiding combination therapy design.