A new model predicts the strength of gene silencing tools

Silscore uses thermodynamics and nucleotide identity to model small interfering RNA potency.

Researchers created Silscore to model the potency of small interfering RNAs (siRNAs). These siRNAs enable programmable gene silencing.

Candidates targeting the same transcript can differ markedly in potency. Silscore is a guide-only model.

The model combines a quadratic preference for overall duplex stability with a position-resolved profile of nearest-neighbour stacking energies. It also uses sparse nucleotide-identity terms.

Duplex stability refers to overall stability. Nearest-neighbour stacking energies are part of a position-resolved profile. Nucleotide-identity terms are selected from residuals left unexplained by thermodynamics.

Silscore uses 26 parameters. This is fewer than half as many as any trainable comparator evaluated here. It attained the highest observed Pearson correlation in three transfer evaluations.

The model shows a preference for weaker seed and stronger central pairing. It also recovers terminal asymmetry through local energy profiles.

Key facts

  • Silscore uses 26 parameters to model siRNA potency.
  • The model attained the highest observed Pearson correlation in the Huesken dataset, Mixset, and Katoh-Suzuki evaluations.
  • Silscore combines thermodynamics with sparse nucleotide-identity terms.

All sections

Edition