Module hpo::similarity

Methods to calculate the Similarity between two terms or sets of terms

Several methods and algorithms to calculate the similarity are already provided in the library, but you can easily add your own as well. The easiest way is to use the Builtins enum.

Examples

Using built-in methods

use hpo::Ontology;
use hpo::similarity::{Builtins, Similarity};
use hpo::term::InformationContentKind;

let ontology = Ontology::from_binary("tests/example.hpo").unwrap();
let term1 = ontology.hpo(12638u32).unwrap();
let term2 = ontology.hpo(100547u32).unwrap();

let ic = Builtins::GraphIc(InformationContentKind::Omim);

let similarity = ic.calculate(&term1, &term2);
println!("The termss {} and {} have a similarity of {}", term1.id(), term2.id(), similarity);
// ==> "The terms HP:0012638 and HP:0100547 have a similarity of 0.2704636"

Create a custom similarity algorithm

Creating you own similarity algorithm is as easy as implementing the Similarity trait.

use hpo::{Ontology, HpoTerm};
use hpo::similarity::Similarity;

struct Foo {}
impl Similarity for Foo {
    /// Calculate similarity based on length of the term names
    fn calculate(&self, a: &HpoTerm, b: &HpoTerm) -> f32 {
        return (a.name().len() - b.name().len()) as f32
    }
}

let ontology = Ontology::from_binary("tests/example.hpo").unwrap();
let term1 = ontology.hpo(12638u32).unwrap();
// ==> "Abnormal nervous system physiology"
let term2 = ontology.hpo(100547u32).unwrap();
// ==> "Abnormal forebrain morphology"

let ic = Foo{};

let similarity = ic.calculate(&term1, &term2);
assert_eq!(similarity, 5.0);

Re-exports

• pub use defaults::Distance;
• pub use defaults::GraphIc;
• pub use defaults::InformationCoefficient;
• pub use defaults::Jc;
• pub use defaults::Lin;
• pub use defaults::Mutation;
• pub use defaults::Relevance;
• pub use defaults::Resnik;

Modules

• defaults
  Contains implementation for some Similarity algorithms that are considered to be default implementations.

Structs

• CachedSimilarity
  Caches the Similarity score for each HpoTerm pair
• GroupSimilarity
  calculate the Similarity score between two HpoSets

Enums

• Builtins
  Contains similarity methods for the standard built-in algorithms
• StandardCombiner
  Default implementations for combining similarity scores of 2 HpoSets

Traits

• Similarity
  Trait for similarity score calculation between 2 HpoTerms
• SimilarityCombiner
  This trait is needed to calculate the similarity between HpoSets.