Genotype to Phenotype Prediction

One-Sentence Definition

Genotype-to-phenotype prediction infers observable microbial traits — resistance, virulence, host range, metabolism — from genome sequence.

Simple Explanation

Read the genome, predict the behaviour. It works well for some traits and poorly for others, and knowing which is which is the skill.

Detailed Scientific Explanation

Two strategies:

  1. Rules/catalogue-based — curated determinant → phenotype mapping; transparent, auditable, limited to known mechanisms (AMR Gene Databases)
  2. Statistical/ML — learn from labeled genome–phenotype pairs; can capture epistasis and unknown determinants, but risks confounding by lineage (Machine Learning Basics for Microbiology)

Where prediction is strong:

  • Mycobacterium tuberculosis resistance to several first-line drugs (mechanisms few and well catalogued)
  • Acquired resistance genes with clean phenotype links (e.g., mecA → methicillin resistance in Staphylococcus aureus)
  • Species and serotype/serovar assignment

Where prediction is weak:

  • Expression-dependent mechanisms — efflux upregulation, porin loss, inducible AmpC (Microbial Transcriptomics)
  • Borderline MICs near breakpoints; heteroresistance; tolerance/persistence
  • Traits shaped by host and environment (biofilm behaviour in vivo, virulence severity)

Additional layers: metabolic phenotype prediction via genome-scale metabolic models (flux balance analysis), growth requirements, and phage susceptibility for phage therapy matching.

Mechanism

Map determinants (genes, alleles, SNPs, copy number, truncations) to phenotype either by curated rule or learned function, ideally with confidence and a “no call” option.

Clinical Importance

  • Genotype-first reporting only works where evidence supports high negative predictive value; otherwise phenotypic testing remains mandatory

Research Importance

  • Systematic discovery of missing determinants when prediction fails

Diagnostic Relevance

  • Basis for sequencing-based susceptibility reporting policies

AMR Relevance

Core aim of clinical microbial genomics; performance judged with very-major-error rules (Model Evaluation in Clinical Microbiology).

Active Recall Questions

  1. Name two traits predicted well and two predicted poorly from genome sequence.
  2. Rules-based vs ML prediction — trade-offs?
  3. Why is heteroresistance a problem for genotype-based reporting?

Connections