Supplemental Data. A Statistical Model for HIV-1 Sequence Classification using the Subtype Analyser (STAR)

R. E. Myers¹, C. V. Gale², A. Harrison³, Y. Takeuchi¹ P. Kellam²†

Supplemental Data

Table 1.

STAR Identifier	Accession	Reference Strain	ORIGIN	Published Subtype
25AAA1A1_36	AF069670	Y	Somalia	A
26AAA1A1_25	AF004885	Y	Kenya	A
28AGAGAG_45	AJ251057	Y	W/C.Africa	A/AG
27A1A1XX_42	M62320	Y	Uganda	A1
51AAA2A2_12	AF286237	Y	Cyprus	A2
52AAA2A2_13	AF286238	Y	D.R. of Congo	A2
33AEAEAE_41	U54771	Y	Thailand	AE
34AEAEXX_41	U51189	Y	Thailand	AE
35AEAEAE_9	U51188	Y	C.A.R.	AE
36AEAEAE_9	AF197340	Y	C.A.R.	AE/E
30AGAGAG_14	AF063223	Y	Djibouti	AG
32AGAGAG_45	AJ251056	Y	W/C.Africa	AG
29AGAGAG_29	L39106	Y	Nigeria	AG
12BBBBBB_0	NC_001802	Y	Coffin,J.M. (Ed.);	B
13BBBBBB_44	U63632	Y	USA	B
14BBBBBB_44	U21135	Y	USA	B
15BBBBBB_22	M17451/ M12508	Y	Haiti	B
59CCCCCC_6	AF110967	Y	Botswana	C
60CCCCCC_23	AF067155	Y	India	C
61CCCCCC_15	U46016	Y	Ethiopia	C
8CDCDCD_40	AF289548	Y	Tanzania	CD
9CDCDCD_40	AF289550	Y	Tanzania	CD
11CDCDCD_40	AF289549	Y	Tanzania	CD
3DDDDDD_0	M22639 (U16633, Z2)	Y	not given	D
5DDDDDD_13	M27323	Y	D.R. of Congo	D
4DDXXDD_13	KO3454/X04414	Y	D.R. of Congo	D
10DDXXDD_42	U88824	Y	Uganda	D
43FFF1F1_7	AF005494	Y	Brazil	F
44FFF1F1_13	AF077336	Y	D.R. of Congo	F
45FFF1F1_25	AF075703	Y	Kenya	F1
46FFF1F1_1	AJ249238	Y	Africa	F1
47F2F2F2_10	AJ249237	Y	Cameroon	F2
49F2F2F2_10	AJ249239	Y	Cameroon	F3
62GGGGGG_25	AF061640	Y	Kenya	G
63GGGGGG_13	AF061642	Y	D.R. of Congo	G
64GGGGGG_13	AF084936	Y	D.R. of Congo	G
65HHHHHH_9	AF005496	Y	C.A.R.	H
68HHHHHH_4	AF190127	Y	Belgium (ori?)	H
69JJJJJJ_13	AF082394	Y	D.R. of Congo	J
70JJJJJJ_13	AF082395	Y	D.R. of Congo	J
1000KK_0	AJ249235	Y	D.R. of Congo	K
1001KK_0	AJ249239	Y	Cameroon	K
1NNNNNN_10	AJ006022	Y	Cameroon	N
2NNNNNN_10	AJ271370	Y	Cameroon	N
536AAAAAA_15	AF071474	N	Ethiopia	A
538AAA1A1_42	AF069671	N	Uganda	A
540AAA1XX_42	AF069673	N	Uganda	A
541AAAAAA_40	AF069669	N	Tanzania	A
542AAAAAA_42	AF069672	N	Uganda	A
500DDCDAA_42	AF075701	N	Uganda	A
569AAA2A2_31	AF286239	N	S. Korea	A2/D
619XXA2A2_31	AF286239	N	S. Korea	A2/D
552AEAEAE_41	AB052995	N	Thailand	AE
553AEAEAE_41	AF259954	N	Thailand	AE
556AEAEAE_33	AY008714	N	S.China	AE
557AEAEAE_33	AY008718	N	S.China	AE
558AEAEAE_41	AB032740	N	Thailand	AE
559AEAEAE_41	AB032741	N	Thailand	AE
560AEAEAE_41	AF197339	N	Thailand	AE
515BBBBAE_41	AF362994	N	Thailand	AE/B
555AEAEAE_41	AF197338	N	Thailand	AE/E
561AEAEAE_9	AF197341	N	C.A.R.	AE/E
544AGAGAG_38	AF107770	N	Sweden	AG
545AGAGAG_10	AF377954	N	Cameroon	AG
546AGAGAG_19	AB049811	N	Ghana	AG
547AGAGAG_10	AF377955	N	Cameroon	AG
548AGAGAG_45	AJ286133	N	W/C.Africa	AG
549Y3AGAG_19	AB052867	N	Ghana	AG
562Y1AGAG_10	AF377957	N	Cameroon	AG
550Y4AGGG_4	AJ276596	N	Belgium (Trans)	AG/C
564Y5AGCC_25	AJ276595	N	Kenya	AG/C
617XXAEAE_41	AF164485	N	Thailand	AGE
535Y1AGY1_5	AJ293865	N	Benin	AGJ
503BBBBBB_3	AF042102	N	Australia	B
504BBBBBB_3	AF042106	N	Australia	B
505BBBBBB_3	AF042104	N	Australia	B
506BBBBBB_3	AF042105	N	Australia	B
507BBBBBB_3	AF042103	N	Australia	B
508BBBBBB_3	AF042100	N	Australia	B
510BBBBBB_28	U23487	N	Manchester, UK	B
512BBBBBB_44	U26942	N	USA	B
514BBBBBB_3	AF042101	N	Australia	B
516BBBBBB_18	AJ271445	N	GB	B
517BBBBXX_39	AF086817	N	Taiwan	B
527BBBBBB_44	U26546	N	USA	B
528BBBBBB_44	AF286365	N	USA	B
529BBBBBB_44	AF049494	N	USA	B
530BBBBBB_44	AF049495	N	USA	B
531BBBBBB_44	AF069140	N	USA	B
571CCCCCC_32	AF286227	N	S.Africa	C
572CCCCCC_6	AF110964	N	Botswana	C
573CCCCXX_6	AF110966	N	Botswana	C
574CCCCCC_6	AF110962	N	Botswana	C
575CCCCCC_6	AF110963	N	Botswana	C
576CCCCCC_6	AF110965	N	Botswana	C
577CCCCCC_6	AF110972	N	Botswana	C
579CCCCXX_6	AF110971	N	Botswana	C
580CCCCCC_6	AF110969	N	Botswana	C
581CCCCCC_46	AF286224	N	Zambia	C
582CCCCCC_40	AF286235	N	Tanzania	C
583CCCCCC_23	AF067154	N	India	C
584CCCCCC_23	AB023804	N	India	C
585CCCCCC_23	AF067158	N	India	C
586CCCCCC_23	AF067157	N	India	C
587CCCCCC_23	AF067159	N	India	C
588CCCCAA_23	AF067156	N	India	C
590CCCCCC_33	AF286231	N	India	C
591CCCCXX_6	AF110968	N	Botswana	C
595CCCCCC_6	AF110961	N	Botswana	C
597CCCCCC_6	AF110973	N	Botswana	C
598CCCCCC_6	AF110974	N	Botswana	C
599CCCCCC_6	AF110975	N	Botswana	C
600CCCCCC_6	AF290028	N	Botswana	C
601CCCCCC_6	AF290030	N	Botswana	C
602CCCCCC_6	AF290027	N	Botswana	C
603CCCCXX_6	AF110979	N	Botswana	C
604CCCCXX_6	AF110980	N	Botswana	C
605CCCCXX_6	AF110981	N	Botswana	C
606CCCCCC_23	AF286232	N	India	C
607CCCCCC_6	AF110976	N	Botswana	C
608CCCCCC_6	AF110978	N	Botswana	C
609CCCCCC_6	AF110977	N	Botswana	C
610CCCCCC_24	AF286233	N	Israel	C
620XXCCCC_6	AF110966	N	Botswana	C
623XXCCCC_6	AF290029	N	Botswana	C
627XXBBF1_7	AF005495	N	Brazil	F
565F2F2F2_10	AF377956	N	Cameroon	F
563Y5AGAG_19	AF184155	N	Ghana	none
509BBBBBB_3	AF146728	N	Australia	none
511BBBBBB_44	AF004394	N	USA	none
513BBBBBB_44	AF070521	N	USA	none
519BBBBBB_37	AF256209	N	Spain	none
520BBBBBB_37	AF256210	N	Spain	none
521BBBBBB_37	AF256204	N	Spain	none
522BBBBBB_37	AF256206	N	Spain	none
523BBBBBB_37	AF256205	N	Spain	none
524BBBBXX_37	AF256211	N	Spain	none
525BBBBBB_37	AF256208	N	Spain	none
526BBBBBB_44	AF075719	N	USA	none
613OOOOOO_35	AJ302646	N	Senegal	O
614OOOOOO_35	AJ302647	N	Senegal	O

Key:

Set of HIV-1 genomes containing Gag, Pol and Env proteins was given an internal accession number of the form; 00AABBCC_11

00 – corresponded to the number of the genome (numbers 1-72 Los Alamos sequences, numbers >500 – sequences derived from GenBank).

AA – Two-letter code used to represent the subtype of the Gag sequence, eg AA = subtype A, AB = subtype A / B recombinant, XX = unassigned.

BB – Two-letter code used to represent the subtype of the Pol sequence.

CC – Two-letter code used to represent the subtype of the Env sequence.

_11 – This value is a key to country of origin, which is detailed within the table.

C.A.R. – Central African Republic

DR Congo – Democratic Republic of Congo

W/C Africa – West Coast of Africa

Equations

STAR score function:

1.1

S - score of query sequence against a subtype alignment.

i-j - amino acid positions from i (start) to j (stop) within query sequence.

fx - frequency of query sequence amino acid at the corresponding position within subtype specific alignment.

fy - frequency of query sequence amino acid at the corresponding position within the all-subtype alignment.

PDT – Positive Discriminant Threshold.

NDT – Negative Discriminant Threshold.

1.2

A score (S) is generated for the query sequence relative to each subtype alignment (n.=11) and each value of the distribution of scores transformed into a Z-score.

S – score.

_S - average of all scores (n = 11).

_S - standard deviation of all scores (n = 11).

1.3

Maximal Z-score (Z_max) for a subtype alignment is used to predict subtype of query sequence.

Hidden Markov Model Analysis

Preliminary attempts to create a subtyping tool compared a very simple PSSM model with the HMMer implementation of Hidden Markov Models. HMMer and the PSSM model performed to a very similar level, misclassifying 1 sequence each. Having established that the PSSM tool could perform as well as the HMM based tool, the HMM based tool was not investigated further. This was for practical considerations, the code for STAR was generated in-house allowing its’ systematics to be adjusted and the scoring function to be deconstructed, allowing analysis along the length of query sequences. HMMer as implemented also proved to be unable to separate the maximal (subtype predictive) scores from other subtype scores, which made it difficult to validate the accuracy of subtype predictions (Figure s1). It should be noted that the utilisation of HMMer could almost certainly have been modified to improve performance. Rather than making a single HMM that modelled the entire HIV-1 Pol (440 aa) sequence alignment, regions of inter- and intra- subtype variation could have been used to create a series of HMM’s that represented the sequence variation along the length of the region. A series of HMMs would also have allowed recombinant sequences to be analysed, assuming that the length of the alignment within the HMM provided sufficient coverage.

Figure s1. The frequency distribution of Z-score from the maximal subtype alignments, (solid line) n. = 141 (141 x 1) and non-maximal subtype alignments (dashed line) n. = 1410 (141 x 10). Data was generated using HMMer to reclassify the 141 sequences used to define the 11 subtype HMMs.