G4 analysis: G4s were called using the Quadparser 2 software  using the default parameters. 


Invariant G4: G4-forming sequences are generally defined as sequences containing four runs of at least three Gs 
separated by one to seven linker nucleotides that forms loops when the triplets of Gs are engaged in Hoogsteen pairing.
Many of the sequences that conform to this definition and that are predicted to form quadruplexes by the Quadparser software 
contain runs of 4 or more Gs which can form G-quadruplex isoforms with different loop length depending on which Gs are involved in Hoogsteen pairing. 
In addition, many predicted G4-forming sequences contain more than 4 consecutive runs of triplets of G’s. All such sequences can form multiple isoforms because 
any combinations of four triplets of Gs can engage in a quadruplex. 
To characterize the effect of loop length we focused on a sub-category of invariant G4 that can only adopt a single conformation. 

Invariant G4-forming sequences were defined as sequences that are smaller than 33 base pairs and that have the following molecular structure: 
G3H1-7G3H1-7G3H1-7G3 (with H equal to A, C or T), or its reverse complement.



Inter-molecular G4: iG4 were called on a Linux workstation using the following regular expression: \(CCC\|GGG\)[A-Z]\{1,7\}\(CCC\|GGG\)[A-Z]\{1,7\}\(CCC\|GGG\)[A-Z]\{1,7\}\(CCC\|GGG\), which finds all the iG4 and the majority of G4s. G4s were then subtracted from the results to obtain a list of all iG4s.