Table 1 The 17 features considered in the models, highlighting the seven robustly selected through stepwise regression (in boldface text). The feature description does not include the scaling performed (Additional file 2: Table S4) to generate more comparable regression coefficients

 Site type

site type

Type of site (8mer, 7mer-m8, or 7mer-A1) [37]

 Site position 9

site9

Identity of nucleotide at position 9 of the site

 Site position 10

site10

Identity of nucleotide at position 10 of the site

 Local AU content

local_AU

AU content within 30 nucleotides of the site [37]

 3′ supplementary pairing

3P_score

Supplementary pairing at the miRNA 3′ end [37]

 Energy of 3′ supplementary pairing

3P_energy

Thermodynamic energy of supplementary pairing at the miRNA 3′ end (ΔG duplex – ΔG seed duplex) (Fig. 3c)

 Predicted structural accessibility

SA

log₁₀(Probability that a 25-nt segment centered on the match to miRNA position 7 is unpaired) (Fig. 3d)

 Probability of conserved targeting^a

P _CT

Probability of site conservation, controlling for dinucleotide evolution and site context [57]

 5′ UTR length

len_5UTR

log₁₀(Length of the 5′ UTR)

 ORF length

len_ORF

log₁₀(Length of the ORF) [51]

 3′ UTR length

len_3UTR

log₁₀(Length of the 3′ UTR) [91]

 5′ UTR AU content

AU_5UTR

Fractional AU content in the 5′ UTR

 ORF AU content

AU_ORF

Fractional AU content in the ORF

 3′ UTR AU content

AU_3UTR

Fractional AU content in the 3′ UTR

 Distance from stop codon

dist_stop

log₁₀(Distance of site from stop codon)

 Minimum distance

min_dist

log₁₀(Minimum distance of site from stop codon or poly(A) cleavage site) [37, 41, 92]

 Weak canonical sites in mRNA

other_sites

Number of 8mer sites in the 5′ UTR and ORF and offset-6mer, 6mer-A1, and 6mer sites in the 3′ UTR [51]