Signature hypotheses raised based on analysis of all within-subject consensus sequences.

Consensus sequences from each subject from all three sets (Table 1, main text) were combined in a hypothesis-raising context (the Test set “All con”). 2 acute signatures were observed (in bold): selecting for a loss of T in acutes at position 415 (discussed in the text), and selecting for F at 721. Key: HXB2 Pos: the HXB2 Env position and amino acid. Aln Pos: The corresponding position in the Env protein alignment. Sig AA: The signature amino acid. Test set: “All con” was based on comparing acute and chronic data using a consensus from each patient and combining all three datasets described in Table 1 in the main text. We raised the q value threshold to 0.5 for this exploratory summary, so we could identify a few potentially interesting sites; only half would be expected to be of interest. “Original” are the six sites for which a signature hypothesis was raised based on the original data; only position 12 H was later supported in the holdout data, so it is discussed further in the main text and was subsequently experimentally validated to regulate expression levels. Here we used our standard q threshold of 0.2. Pattern: “A to !A” means the signature amino acid is predicted in the maximum likelihood tree to be A in the most recent ancestral node of the subject, but to have changed to not being the signature amino acid (“!A” means “not A”) in the subject. This change contrasted to the signature amino acid remaining the same in the contingency table (The signature amino acid A it found in the recent ancestor and the leaf node). “!A to A” is the inverse situation where the ancestral state is not the signature amino acid. FS: Fiebig Stage.