The bHLH domain has been categorized into 5 major classes (Atchley and Fitch 1997). The bHLH288 dataset contains 288 amino acid sequences with samples from each class. While the basic and helix regions have well defined structures consisting of 13 and 15 amino acids respectively, the loop region has variable length. To prevent gaps, the loop was truncated in some proteins so only 51 sites are retained and partitioned into basic (1-13), helix(14-28), loop (29-36), and helix(37-51) regions.
Atchley et al 2005 performed factor analysis on a set of Amino Acid Indices (AA54) and inferred a 5 factor latent variable structure relating amino acid characteristics using SAS. An equivalent analysis was performed using factor.pa.ginv from the HDMD package in R. Based on the relationship between factors and variable descriptions, the latent variables are defined as Factor1 (PAH): Polarity, Accessibility, Hydrophobicity; Factor2 (PSS): Propensity for Secondary Structure; Factor3 (MS) : Molecular Size; Factor4 (CC): Codon Composition; Factor5 (EC): Electrostatic Charge. While the Factor Analysis loadings were the same, R and SAS calculated scores slightly differently. AAMetric are scores from the R factor analysis which convey the similarities and differences among amino acids (rows) for each latent variable (columns).
Atchley et al 2005 performed factor analysis on a set of Amino Acid Indices (AA54) and inferred a 5 factor latent variable structure relating amino acid characteristics using SAS. Based on the relationship between factors and variable descriptions, the latent variables are defined as Factor1 (PAH): Polarity, Accessibility, Hydrophobicity; Factor2 (PSS): Propensity for Secondary Structure; Factor3 (MS) : Molecular Size; Factor4 (CC): Codon Composition; Factor5 (EC): Electrostatic Charge. AAMetric.Atchley are scores from the factor analysis which convey the similarities and differences among amino acids (rows) for each latent variable (columns).
From approximately 500 indices listed in www.genome.jp/aaindex as of 2005, a sample of 54 Amino Acid Indices were selected to represent the range of structural and functional attributes. Each index was normalized to have mean 0 and variation 1.
providing 
.