1
4
Revealing the recent demographic history
of Europe via haplotype sharing in the UK
Biobank.
5
Supplementary Information Appendix
2
3
6
7
8
Authors
Edmund Gilbert1,2*, Ashwini Shanmugam1,2,3, Gianpiero L. Cavalleri1,2,3*.
9
10
11
12
Affiliations
13
* Corresponding Author
14
15
Table of Contents
16
Affiliations ............................................................................................................................................... 1
17
Supplementary Data 1 - UK Biobank Self-Reported Ethnicity ................................................................ 3
18
Supplementary Data 2 – Individual Country Population Structure ........................................................ 4
19
Supplementary Data 2.1 - Per Country PCA Distributions .................................................................. 4
20
Supplementary Data 2.2 - Principal Component Analysis of UK Biobank Europeans ...................... 12
21
Supplementary Data 2.3 - Per ADMIXTURE Component Proportions .............................................. 13
22
Supplementary Data 2.3 - Comparison to Human Origins References............................................. 14
23
Supplementary Data 3 – Comparison on Linked and Unlinked Methods............................................. 15
24
Supplementary Data 4 - Genetic Structure of European Leiden Clusters ............................................ 18
25
Supplementary Data 5 – Detailed European Genetic Landscape ......................................................... 22
26
North-Western Europe ..................................................................................................................... 22
27
Central-Eastern Europe ..................................................................................................................... 23
28
Southern Europe ............................................................................................................................... 24
29
Supplementary Data 6 - Malta .............................................................................................................. 25
30
Supplementary Data 7 - Mixed Europeans ........................................................................................... 26
31
Supplementary Data 8 - IBD Ne Curves ................................................................................................ 31
32
Supplementary Data 9 - Spain .............................................................................................................. 36
1. School of Pharmacy and Biomolecular Sciences, Royal College of Surgeons in Ireland, Dublin.
2. FutureNeuro SFI Research Centre, Royal College of Surgeons in Ireland, Dublin.
3. The SFI Centre for Research Training in Genomics Data Science, Ireland.
Authors.................................................................................................................................................... 1
33
34
35
References ............................................................................................................................................ 39
36
37
38
39
40
41
42
43
44
Supplementary Data 1 - UK Biobank Self-Reported Ethnicity
Whilst sampling European ancestry within the UK Biobank1, we selected based on European birthplace
and self-reported ethnicity, selecting the background which fell under the parent description “White”
(F.21000 = 1/1001/1002/1003) or “other” (F.21000 = 6) (see Methods). With comparison to West
Eurasian references from the Human Origins2 by projection principal component analysis (PCA) this
reveals a sample of European ancestry across the continent (SI Appendix Figure S2.11). We further
recorded the proportions of the selected self-reported ethnicity categorises within each sampled
country/region of birth (Supplementary Figure 1.1), as well as each Leiden3 cluster subsequently
detected (Supplementary Figure 1.2).
Brit. & Ire.
W Europe
N Europe
C Europe
E Europe
F.21000.Label
0.75
White
White − British
0.50
White − Irish
White − Other
0.25
46
Channel Islands
England
Isle of Man
Northern Ireland
Orkney
Republic of Ireland
Scotland
Wales
Belgium
France
Netherlands
Denmark
Faroe Islands
Finland
Iceland
Norway
Sweden
Austria
Czech Republic
Germany
Hungary
Poland
Slovakia
Switzerland
Belarus
Estonia
Latvia
Lithuania
Russia
Ukraine
Gibraltar
Italy
Malta
Portugal
Spain
Albania
Bosnia and Herzegovina
Bulgaria
Croatia
Greece
Kosovo
North Macedonia
Romania
Serbia
Serbia/Montenegro
Slovenia
Other
0.00
47
48
49
S Europe
1.00
Cyprus
Turkey
Proportion Membership
Near East
SE Europe
45
Supplementary Figure 1.1 – Per-country/region of birth proportions for five self-reported ethnicity
categories from the UK Biobank. The ethnicity category corresponds to UK Biobank phenotype code
F.21000.
50
White
White − British
0.50
52
53
54
White − Irish
White − Other
Other
Malta
Mixed European
Spain
France & Switz.
Cyprus
Portugal
Turkey
Albania & Greece
Italy
Greece
Finland
Mixed Scand.
Russia
Latvia & Lithuania
Poland
NW Balkans
NE Balkans
CE Europe 1
CE Europe 2
Wales 2
Channel Islands
Wales 1
France
England 2
England 1
Isle of Man
Ireland 2
Ireland & N.Ireland
Orkney
Ireland 1
N.Ireland & Scotland 4
N.Ireland & Scotland 2
N.Ireland & Scotland 3
N.Ireland & Scotland 1
Norway
Sweden 1
Denmark
Belgium
Switzerland
Netherlands
0.25
0.00
51
F.21000.Label
0.75
Belgium & France
Proportion Membership
1.00
Supplementary Figure 1.2 – Per-Leiden-cluster proportions for five self-reported ethnicity categories
from the UK Biobank. The ethnicity category corresponds to UK Biobank phenotype code F.21000.
55
Supplementary Data 2 – Individual Country Population Structure
56
57
58
59
60
61
62
Supplementary Data 2.1 - Per Country PCA Distributions
In sampling of UK Biobank1 participants with European ancestry and birthplace we performed an initial
principal component analysis (PCA), investigating the genetic structure in the sample of 5,500
individuals, and the genetic structure sampled in each individual country/region of birth. Below we
record the distribution of individuals for each such region, grouped by geographic proximity. Each plot
shows the coordinates from the PCA shown in Figure 1, highlighting individuals from each individual
region in red and every other individual as a small grey circular point.
63
64
65
66
Supplementary Figure 2.1 – Principal Component coordinates of UK Biobank individuals with Cypriot,
or Turkish birthplace.
67
68
69
70
Supplementary Figure 2.2 – Principal Component coordinates of UK Biobank individuals with Greek,
Kosovan, North Macedonian, Romanian, Serbian/Montenegro, or Slovenian birthplace.
71
72
73
Supplementary Figure 2.3 – Principal Component coordinates of UK Biobank individuals with
Gibraltarian, Italian, Maltese, Portuguese, or Spanish birthplace.
74
75
76
Supplementary Figure 2.4 – Principal Component coordinates of UK Biobank individuals with
Belarusian, Estonian, Latvian, Lithuanian, Russian, or Ukrainian birthplace.
77
78
79
Supplementary Figure 2.5 – Principal Component coordinates of UK Biobank individuals with Austrian,
Czech, German, Hungarian, Polish, Slovakian, or Swiss birthplace.
80
81
82
83
Supplementary Figure 2.6 – Principal Component coordinates of UK Biobank individuals with Danish,
Faroese, Finnish, Icelandic, Norwegian, or Swedish birthplace.
84
85
86
Supplementary Figure 2.7 – Principal Component coordinates of UK Biobank individuals with Belgian,
French, or Dutch birthplace.
87
88
89
90
Supplementary Figure 2.8 – Principal Component coordinates of UK Biobank individuals with British
or Irish birthplace.
91
92
93
94
95
96
Supplementary Data 2.2 - Principal Component Analysis of UK Biobank Europeans
To supplement Figure 1, we show below the additional principal components three through to ten.
These were calculated using the same methodology as those principal components shown in Figure 1,
with the same legend and codes for country/region of birth. For each of the four panels, the x-axis
principal component is given first and then the y-axis principal component (PC), i.e., for “PC3-PC4”, PC
three forms the x-axis and PC four forms the y.
97
98
99
100
101
Supplementary Figure 2.9 - Principal component (PC) analysis of 5,500 Europeans from the UK
Biobank1 using PLINK4,5 across PCs three through ten.
102
103
104
105
106
107
108
Supplementary Data 2.3 - Per ADMIXTURE Component Proportions
Further, we performed ADMIXTURE6 analysis on the same individuals and common markers there
were used for the PCA using PLINK4,5 above (see Methods). ADMIXTURE v1.3 is a fast modelled-based
maximum likelihood estimation of an individuals’ ancestry, assuming k ancestral “populations” or
components. To explore the genetic structure in the UKBB 5,500 sample we ran ADMIXTURE over k
values from two to seven, run ten replications for each k value and choosing the replicate with the
high log-likelihood and lowest cross-error validation score. We show the results below:
109
110
111
112
113
114
Supplementary Figure 2.10 - ADMIXTURE ancestry components for each sampled European
country/region of birth over k values 2 to 7. Populations are ordered into the same groups of
geographically adjacent regions as in Supplementary Figures 2.1-8.
115
116
Supplementary Data 2.3 - Comparison to Human Origins References
A
B
HO
Population
0.00
−0.05
117
118
119
120
121
122
123
124
125
0.00
0.03
Principal Component 1
0.06
Jew_Turkish
Adygei
Jew_Yemenite
Balkar
Albanian
Chechen
Bulgarian
Georgian
Croatian
Iranian
Greek
Iranian_Bandari
Romanian
Kumyk
Basque
Lezgin
Italian_North
Nogai
Italian_South
North_Ossetian
Maltese
Armenian
Sardinian
Assyrian
Sicilian
Cypriot
Spanish
Druze
Spanish_North
Jordanian
English
Lebanese
French
Lebanese_Christian
Orcadian
Lebanese_Muslim
Scottish
Palestinian
Belarusian
Syrian
Czech
Turkish
Estonian
BedouinA
Hungarian
BedouinB
Ukrainian
Saudi
Chuvash
Yemeni
Finnish
Libyan
Icelandic
Moroccan
Lithuanian
Jew_Ashkenazi
Mordovian
Jew_Georgian
Norwegian
Jew_Iranian
Russian
Jew_iraqi
Saami_WGA
DK
RU
0.05
RU
TR
RU RU
Principal Component 2
Principal Component 2
0.05
Abkhasian
0.00
RU
TRTR
GR
NO
RU
TR
TR TR
RUNL
RU
UA
TRTR
TR
FI NO
TR
TRTR
TR
CZ
UA
TR
SE LT RU NLFIDE
TR
TR
TR
TR
TR
TR
TR TR
TR
TR
TR
RO
TR
TR
FI
FI
FI RU
RU LT
TR
TR
TR
FI
FIFI
TR TR
TR
FI
TR
FI
FI
FI
FI
FI
TR
TR
RU
RU
TR
TR
RU
FI
FI
RU
TR
FIRU
RURU RU
TRTR
FI
TRTR
TR
TR
TR
FI
TR
FI
FI
RU
NO
FI
FI
FI
FI
RU
FI
TR
FI
LT
TR
TR
FI
FI
FI
TR HU
TR
FI
NO
FI
FI
RU
TR TR TR
FI
EE
LV
TR
RURU
FI
FI
RU
NO
FI
FI
TR
FI
TR
FI
RU
RU
RU
TR
FI
NO
FI
CZ
RU
FI
RU
TR
FI
LT
FI
RU
FI
UA
FI
RU
LV
RU
LT
RU
FI
RU
EE
FI
HU TR
FI
BG TR TR
FI
LT RU
TR GR
LT
TR
SE
RU
FI
FI
FI
RU
FI
RU
LV
RU
FI
TR
RU
LT
RU
FI
RU
FI
RU
RU
FI
FI
FI
FI
RU
TR
FI
RU
TR
EE
FI
RU
TR
DE
LV
LT
FI
LV
FI
UA
LV
UA
LV
FI
FI
FI
EE
LV
LT
RU
FI
CZ
LV
RU
FI
RU
PL
FI
LT
LT
LT
RU
UA
FI
RU
TR
RU
RU
TR
FI
LT
RU
LV
LV
LV
FI
LT
RU
FI
LT
RU
FI
FI
TR
TR TR
TR
RU
RU
LT
FI
EE
UA
RU
SE
RU
TR
LT
LV
UA
EE
TR
LT
RU
PL
RU
RU
PL
RU
LT
TR
FI
FI
EE
TR
FI
BG
LT
RU
PL
FI
LT
FI
RU
LT
LV
RU
RU
RU
NO
LT
HU
SE
LV
LV
PL
LT
RU
LV
LV
RU
EE
UA
GR
RU
LV
LT
FI
LV
TR
RU
FI
LT
PL
LT
LT
PL
RU
TR
TR
RO
LV
LV
FI
RU
BY
LT
PL
LT
LV
LV
LT
LT
LT
UA
FI
BY
TR
RU
LT
DK
UA
UA
PL
UA
PL
CY TR
FI
LV
FI
NO
RU
TR
EE
UA
FI
RU
RU
LV
LV
RU RU TR
LV
PL
DE
RU
RU
RU
LV
EE
RU
RU
UA
LV
FI
LT
FI
RO
UA
LV
FI
UA
PL
LT
PL
UAUA
R
U
PL
LT
FI
FI
PL
RU
LT
PL
LT
RU
MT
LT
SE
LV
LV
PL
LV
UA
PL
PL
LV
EE
LV
NO
PL
LT
CZ
CY
AT
UA
PL
HU
FI
UA
EE
PL
RU
BGTR
LV
LTLT
RU
PL
PL
PL
PL
RU
LT
PL
LV
TR
GR
LT
FI
PL
BA
LV
LV
LT
RU
PL
UA
NO
FI
LV
RU
UA
SE
RU
RU
LT
EE
BY
UA
UA
TR
PL
PL
LT
CZ
PL
UA
PL
RU
LV
LT
RU
HR
PL
NO
BY
DE
RU
LV
PL
RU
PL
RU
UA
LT
RO
AT
TR
PL
GR
UA
PL
PL
HU
GR
LV
LV
HU
HU
RU
TR
HU
UA
PL
UA
PL
CZ
PL
UA
PL
RU
CZ
PL
DE
PL
CZ
HR
FI
PL
LT
LT
PL
FI
UA
PL
SE
NO
UA
GR
PL
PL
TRTR
LT
LT
HU
RO
UA
LT
FI
UA
LTLV
UA
LT
PL
SE
SE
PL
FI
SK
HU
SE
PL
PL
DE
EN
CZ
LV
FI
LV
SK
PL
HU
PL
HU
RO
MT
RU
CZ
NL
SE
PL
PL
SK
BG
PL
PL
UA
HU
BA
RU
BG
AT
TR
DE
CZ
RO TR
UA
PL
CZ
DE
PL
PL
CZ
PL
SK
PL
SE
EN
HU
PL
SE
PL
UA
UA
HU
CZ
AT
GR
SE
RU
SE
HU
PL
PL
HU
BA
SE
PL
SE
CH
SK
SC
NL
LT
PL
DK
NO
PL
BG
UA
PL
DE
UA
PL
SK
SE
RO
PL
SK
CZ
LV
HU
FI
AT
PL
SE
PL
CZ
HR
PL
GR
LT
SE
NL
SE
PL
HU
SE
DE
RO
RU
RO
SE
NO
SE
SC
PL
SK
HU
PL
BG
NI
RS
BG
SE
BA
HU
HU
PL
RO
PL
SK
AT
GR
IE
SE
NL
NO
AT
PL
NI
SE
PL
FI
SE
HR
RO
CZ
CZ
CZ
MK
PL
HU
BG
SE
FR
DK
BA
SE
SE
GRGRCY
LT
NO
PL
BA
NO
SE
CZ
MK
BA
BA
RO
UA
RO
SK
NO
HU
RO
BG
GR
IE
BG
IS
DE
NL
PL
FR
SE
AT
DK
DK
NO
CZ
AT
SK
CZ
FR
DE
SK
AT
NL
RO
FI
PL
RS
PL
DK
PL
RU
HU
PL
NI
AT
RU
HU
CY
HU
BA
CZ
HU
RS
NO
HU
BE
HU
RS
DE
CZ
RU
NO
RS
BG
NO
PL
PL
NO
CZ
CZ
IM
AT
RO
CZ
CZ
SE
PL
PL
SK
AT
HU
HR
DE
NO
NL
RS
RO
HU
CZ
DE
RU
NO
PL
NI
PL
SE
AT
DE
AT
CY
SC
DK
BA
CY
CZ
SE
CZ
SK
DK
HR
HU
RO
BG
SE
AT
SE
PL
DE
HU
NL
CZ
SK
PL
BG
GR
CZ
CZ
SE
RU
GR
PL
PL
HU
BG
DK
FI
SC
MT
BA
NO
DK
PL
BA
TR
SE
SE
RO
CY
DK
PL
IE
PL
HR
NL
SK
HR
BA
RS
SE
NL
DK
AT
IE
DK
PL
CZ
SC
BG
CZ
IM
HR
HR
BG
NO
SE
BG
BG
AT
NI
IE
SK
IE
AT
HU
SE
RS
DE
AT
HU
NO
NO
NO
GR GR GR
NO
DK
NL
CZ
PL
PL
SE
DE
SE
CZ
RO
HU
RO
BG
RU
BG
NO
SK
RO
BA
TR
HU
AT
RO
NO
NO
NO
CZ
IT
TR
SE
NL
SK
CZ
SK
AT
MK
SE
CH
HU
MT
NI
NO
IS
RO
SI
MK
HU
HR
PL
IE
SK
HU
RO
GR
BG
IE
SI
SE
BG
RO
LV
BG
HR
NO
SE
CZ
AT
CZ
DK
PL
CH
GI
CY
SE
NO
SE
IE
SC
DE
DK
SE
DE
BG
GR GR
NO
EN
NI
NO
CI
PL
RS
DE
HU
BA
RO
GR
DK
DK
EN
EN
RO
CZ
AT
RO
SE
IM
DE
SI
NL
DE
CZ
SC
CY
PL
HR
BA
RO
BG
CY
NL
SE
SE
EN
CZ
HU
RO
GR
NO
NL
RS
EN
BE
IE
HU
RS
CY
NO
DK
SI
RS
SE
DK
SC
HU
NO
AT
HU
SE
DK
CZ
AT
IE
RO
NO
DK
RS
HR
BG
DE
SE
HU
HR
RO
CZ
SE
CY
OK
HU
SE
AT
NL
NI
DK
HR
IM
SC
HU
AT
RU
NI
IM
PL
DE
RU
FO
IS
NO
WA
SC
IE
SI
BA
PL
AT
BA
SE
DE
BG
CY
PL
SK
BG
BG
RO
GR
TR
SE
IT
IS
CZ
BG
IS
PL
DK
CZ
AT
SI
RS
CY
WA
PL
IE
NI
AT
PL
AT
RS
RS
DK
AT
DE
WA
DE
SE
NL
DE
DK
SE
NL
CZ
EN
BG
BG
DK
GI
WA
SE
IE
IM
DE
IT
CY
CZ
AT
BA
NL
BA
NO
MT
BG
SE
NL
SK
RS
BG
IE
EN
AT
RO
DK
WA
IM
DK
HU
HR
RO
IM
NI
MT
CY
EN
BA
HR
NI
BA
RO
RO
OK
AT
HU
NO
NI
CI
MT
AT
AT
DK
BG
BG
CY
EN
DE
RU
NO
SE
NI
AT
RS
AT
MK
WA
DK
SK
RU
SE
CZ
HU
HR
PL
BG
RO
IE
NI
CY
IE
SE
AT
IS
EN
NL
BE
SC
GI
PL
BE
CZ
CY
IS
CZ
PL
HU
NI
IS
DK
NL
FI
NO
CZ
CZ
BE
SE
RO
FR
NO
DE
NI
EN
RO
HR
PL
DK
DK
SE
SC
NL
NO
DE
SC
HR
MT
CZ
IE
CZ
HU
RS
SE
DK
CI
SI
NL
DK
NO
IE
BA
SE
OK
CH
WA
WA
SC
HU
RS
BG
DK
OK
EN
NL
BG
SE
SE
EN
WA
CZ
WA
NI
EN
DE
SK
RO
DK
AT
MT
HU
RS
GR
SE
SC
WA
MT
NI
WA
GR
SC
NO
SC
MK
IE
SE
CI
IM
NI
EN
HR
BG
SE
AT
RO
WA
EN
PL
WA
AT
RS
IE
AT
DK
GI
DE
GR
DK
NL
RS
GR
DK
NO
FR
SE
DK
SC
BG
GR
EN
MT
BE
CY
DE
DE
IT
AT
HR
HU
IS
WA
BE
DE
RU
GR
GR
SC
SC
AT
NO
NI
SE
EN
IM
TR
AT
GR
DK
SK
CZ
EN
IE
WA
CH
RS
BG
MT
CZ
BG
SE
SE
NI
MT
NI
OK
NL
RO
DE
HU
TR
SE
NL
NI
TR
GR
OK
DK
IE
DK
NI
OK
NO
AT
IE
WA
DE
PL
PL
HU
RS
RO
IE
SE
IE
MT
BA
DK
NO
SC
SE
IM
DK
IE
BA
CH
DE
DK
SC
EN
DE
NL
SC
IM
WA
AT
SE
BG
NI
HR
CY
SE
IS
NI
AT
EN
HR
CY
NO
SE
DK
BE
NO
NI
BA
DK
NO
NI
NL
SC
RS
MK
BG
GR
NL
NL
DK
PL
RS
DK
NL
RS
BG
GR
IS
CI
NL
AT
CY
CH
NO
DE
DE
RO
BG
IWA
E
CY
WA
DE
HU
CH
CY
NI
CI
DE
GR
GR
SE
CI
NO
CZ
NL
FR
BG
HU
NI
NI
FI
GR
RO
EN
OK
EN
BG
EN
NL
AT
DK
DE
WA
AT
RO
NO
EN
NO
SC
CH
KO
SE
NI
SC
WA
AT
CY
AT
BG
FR
MT
CY
DK
NI
NL
NL
DE
BA
CZ
WA
DK
IM
NL
CZ
RS
BG
DK
PL
WA
AT
MT
BA
RO
WA
DK
BE
CY
WA
CY
SC
SC
AL
SC
NI
IM
SE
IE
MT
IE
SC
NO
GR
DE
NI
HU
DE
WA
AT
IE
NI
RS
HR
CY
BE
BA
CI
IE
SC
OK
IE
EN
UA
GR
SE
NO
SE
IE
DK
MT
NI
RS
RS
MK
WA
EN
IE
FR
CH
MT
DK
IS
EN
DE
SE
NO
CY
IE
EN
CZ
AT
IM
CY
CY
EN
HR
CYCY
CY
WA
SC
CZ
NL
AT
CZ
SE
DE
DK
MT
CZ
RS
AT
OK
OK
NL
NI
HR
BA
BG
DK
DE
NI
HU
SE
IM
DK
TR
FO
NL
GI
CI
RO
DK
GI
NI
FR
NL
SI
CY
DK
SC
KO
GR
GR
GR
AT
AT
HU
IEIE
CH
CH
BG
GR
CI
UA
IE
IE
PT
BE
NI
DK
CY
DE
NL
NL
CY
IE
FR
CY
CI
CY
CH
EN
MK
GI
NI
GI
WA
MT
NI
SC
HR
RS
BG
IE
NO
MT
IE
SC
GR
CY
SE
MT
DE
NL
AT
DK
DK
DK
SC
DK
FO
NL
CZ
EN
DE
NL
NI
NO
NL
DK
GI
DK
CH
CY
NO
GR
NI
CI
EN
WA
WA
AT
HR
DK
WA
SC
IS
DK
SE
PL
CZ
BG
DK
EN
MT
SE
TR
CI
DE
CY
RO
IE
NO
WA
DE
MT
CY
NO
IE
NI
IE
NI
DE
IE
HU
SE
SE
EN
DE
SC
CZ
HR
NI
BE
IE
BE
AT
RS
GR
SE
BG
CH
NL
AT
AT
IS
DK
IE
DK
SE
SC
BG
NO
EN
OK
NL
EN
BE
IM
SC
BG
CZ
IS
OK
DK
SE
WA
IE
IE
DKGI
CI
AT
CH
BA
DE
NI
CI
BG
CY
TR
MT
MT
CI
CY
IS
FR
WA
SC
SC
HR
CH
DE
ENDK
CY
MT
SE
NI
CI
WA
EN
NI
NL
AT
CY
WA
DK
GI
DE
UA
DK
MT
IE
DE
GR
GR
CY
SE
NI
SC
DE
CH
AT
CY
NL
GI
DE
NL
WA
SE
SC
CY
AT
CH
AT
NL
CI
EN
MT
AT
AT
NO
SE
DK
SC
IE
NI
NL
CZ
CZ
SC
NL
CY
GI
AT
RS
GR
EN
BE
MT
BE
CH
OK
MT
NI
DE
EN
NO
MT
BE
NL
OK
OK
CH
TR
GR
DK
ES
IE
GR
GR
NL
IM
CY
WA
CH
NL
NI
DE
EN
AT
HU
NO
CY
DK
SC
BE
GR
DE
IE
PT
EN
WA
BE
BE
AT
BE
AT
GR
DK
BE
BE
WA
HU
DE
NI
MT
NL
BG
CY
CY
IE
NL
CY
BE
DK
GI
AT
BE
IM
IM
EN
SC
DK
SC
NI
IT
NO
SE
IE
SC
AT
SC
CI
AT
MT
AT
WA
EN
WA
CI
EN
EN
NL
HR
AT
DE
BE
DK
WA
IM
CI
AT
NI
DE
CH
IS
WA
MT
CI
CI
DK
IM
IE
EN
GR
DE
CY
WA
NI
NO
EN
DE
SE
CZ
HU
OK
PL
MT
NI
CY
GR
NL
NI
CI
MK
MT
DK
NI
EN
DE
ES
DE
AT
DK
SE
NL
FR
CH
MT
IT
SE
IE
CH
NI
MT
IE
CY
NI
NL
CY
SE
WA
NO
IE
DK
CI
SC
HU
FR
SC
SE
IE
FR
BE
IE
EN
CI
GI
EN
SC
NI
SC
CI
AT
CY
MT
SE
SC
NI
IE
BE
BE
MT
NL
IE
NO
NL
DE
NO
CH
SC
SC
DE
EN
CH
IT
GR
NO
NO
NL
GI
IM
AT
CI
CI
C
Y
AT
HR
IE
WA
CH
GR
DE
DE
RS
NL
IM
MT
CY
FR
N
L
RU
AT
HU
NI
WACI
BE
BA
OK
FR
BE
EN
CI
HR
IM
NI
PT
IM
MT
HR
DE
ES
AL
MT
CH
CI
AT
AT
NL
BE
NL
CH
BE
BE
SE
SC
CY
SC
AT
BE
WA
GI
PL
CY
DK
SC
BE
BE
MT
BE
NL
NO
DE
NL
MT
MT
CZ
NL
GR
NI
SE
WA
CY
IE
GR
GR
EN
CI
CY
DE
CH
CH
WA
NI
BG
IE
NL
AT
RU
SC
EN
IM
IM
AT
AT
AT
NL
IT
GR
IE
DK
IM
F
NI
DE
R
IT
CI
AT
KO
DE
WA
CZ
MT
EN
CI
OK
BE
BA
DE
NO
NL
SC
CI
NI
MT
AT
AT
NO
IE
EN
SC
CY
NL
BE
IE
OK
WA
MT
NI
DK
DE
WA
DE
AT
AT
NL
NI
IE
TR
IE
CH
CH
OK
SE
SC
FR
NI
CY
MT
BE
IE
DE
SC
NI
EN
HU
NI
MT
CY
IE
SC
SC
CH
WA
IM
WA
GI
EN
FR
DE
IE
WA
NI
IE
EN
CH
FI
EN
WA
WA
MT
DK
SC
MT
GI
FR
BE
SE
FR
OK
IE
GI
CI
NI
CI
WA
DE
AL
IE
CI
DK
EN
CY
BG
CY
NI
EN
DE
AT
MT
N
IDE
BE
HR
MT
NL
AT
FR
CI
DK
FO
NL
NI
BE
AT
IT
CY
EN
MT
NL
CH
AT
PL
NI
IE
NI
SC
EN
WA
CI
HR
IT
GR
NO
EN
GI
BE
CI
GR
GR
WA
SC
IT
RS
DK
GR
DK
IE
NL
DE
IM
FR
IT
NO
OK
IM
BE
NO
EN
CH
AT
RO
MK
BE
SC
CY
FR
CH
IT
BE
FR
DE
AT
CI
IT
SC
EN
CH
WA
IT
CH
SC
AT
FR
MT
AT
RS
DK
EN
MT
FO
OK
EN
IM
SC
NL
BE
CI
FR
GR
IM
GI
SC
CY
CY
DE
SC
NI
GR
WA
IM
DE
FR
WA
RO
IT
CY
CY
IE
SC
DK
BE
NL
AT
EN
DK
GI
DE
BE
IT
AL
KO
HU
IE
IM
SC
CH
FR
NL
NL
DE
MT
IE
GI
CY
CY
WA
CI
CY
AT
SC
BE
EN
BE
CY
GI
NL
CY
MT
EN
BE
AT
GR
CH
EN
GI
DE
HR
CH
IT
GR
SE
WA
IE
EN
DE
HU
SC
FR
FI
GR
GR
GR
GR
NO
BE
FR
DE
DE
BE
AT
PT
WA
NL
CI
NI
PT
BE
IT
WA
RS
GR
DE
NO
DK
SE
IE
OK
CY
DE
GR
IT
EN
NL
BE
CYCY
IE
SE
NI
SC
NI
NL
NL
NI
AT
CH
GR
GR
WA
BE
CY
WA
CI
SC
AT
E
FR
N
FR
EN
CI
SE
NI
CY
GI
CY
IE
CI
IE
EN
IT
CI
NL
AT
FR
NL
SC
CH
PT
IM
BE
BE
CH
CZ
DE
NO
EN
EN
CI
DK
NL
CH
EN
DE
IE
FR
CH
DK
NL
EN
DE
CH
NL
CH
IT
BG
MK
DE
BE
IM
TR
CZ
BY
IM
AT
BE
NLBE
DE
SC
DE
CH
EN
CY
SC
WA
WA
SC
WA
SC
EN
IE
BE
EN
MT
BE
CY
NI
NO
WA
SE
SC
CI
AT
NL
NO
CY
MT
NL
CH
EN
AL
IE
MT
RS
KO
SC
EN
AT
CY
WA
WA
WA
NO
WA
OK
NI
CH
NL
NI
EN
GR
MT
NI
AT
AT
IE
FR
MT
MT
HU
WA
BE
IT
DE
NL
GI
GI
FR
FR
AT
IT
RS
IT
DK
IE
WA
NI
FR
AT
EN
NL
EN
BE
GR
IT
NI
CH
GR
NI
MT
CY
WA
NI
NL
FR
CH
GR
NL
MT
SC
CY
NI
WA
AT
TR
CH
WA
WA
DE
GR
GR
CY
FR
MT
GR
BY
IT
DK
WA
AT
FR
EN
BE
IT
WA
BE
DE
FR
CY
CY
GI
BE
EN
CI
AT
IT
DE
NL
NL
AT
CH
CI
NL
MT
IT
CY
CI
CI
FR
CH
SC
MK
GR
HU
CY
FR
DE
TR
KO
GR
GR
CI
IT
AT
WA
SC
CY
NL
MT
DK
FR
IT
DK
NI
CH
FR
CY
FO
BE
NL
GR
GR
AT
CY
SE
MT
IE
OK
IE
NI
FR
IT
BE
AT
CY
NL
MT
IT
RU
CY
MT
IE
IE
CI
SC
NI
AT
DE
HU
PL
SC
FR
CH
CH
IT
IT
EN
CH
CI
FR
IT
IE
NI
CI
NI
MT
CH
CI
IT
CY
DKCI
OK
IE
SC
EN
DE
WA
DE
IT
NL
SC
FR
IT
DE
WA
NI
BE
CH
NL
W
WA
A
IT
GR
MT
IM
WA
WA
CI
MK
HU
NI
EN
SC
DK
EN
GR
BE
IT
CY
SC
NO
CY
BE
CH
BE
CH
AL
DK
GR
NL
OK
SC
CH
CY
IM
WA
SC
NI
WA
CH
IS
WA
EN
NL
GI
ITFR
GR
OK
FR
SE
DE
SC
FR
MT
CH
RU
CY
MT
NL
DE
BE
OK
WA
DE
AT
AT
IT
DK
GI
FR
CH
MT
MK
GR
RO
NL
NI
BE
CY
MT
CI
CI
RU
IM
FR
BE
IT
IT
CY
EN
DE
SE
CY
CH
CY
CZ
WA
FR
TR
CH
AT
CY
SC
CY
BE
IM
CI
IE
AL
RU
NL
IE
CI
DE
SC
FR
FR
FR
AT
FR
BE
MT
IT
MK
GI
CY
IT
EN
NI
MT
WA
CH
GR
GR
DE
AT
IE
RS
AL
IT
NI
CI
CI
CI
CY
CY
DE
IT
FR
CH
MT
PT
MT
FR
IT
MT
BE
CH
IT
UA
FR
WA
NI
CH
BE
FR
DE
EN
HR
BE
IE
NI
CH
MT
WA
DE
PT
CH
FR
WA
NL
CH
CY
BE
AT
NL
SC
IT
CY
DE
DE
FR
IM
WA
CI
IM
SE
GR
GR
GRCY
CY
CI
CH
FR
IT
SC
CY
IT
BE
IM
EN
DE
DK
IE
BE
CH
IT
IT
CY
NI
IE
MT
WA
SC
WA
IM
GR
SC
BE
NI
CH
DE
FR
EN
DE
IT
MT
CY
SC
CY
DE
IT
IT
EN
AT
CH
NL
CY
SC
CI
BE
WA
CI
FR
BE
CI
FR
DE
MT
FR
IT
WA
FR
CH
FR
FR
IT
IM
EN
IT
CY
FR
AT
BE
IT
IT
GR
UA
IE
EN
BE
BE
CH
IT
NL
EN
IT
HU
NL
BE
IT
MT
GI
BE
CH
FR
CH
BE
CY
IM
EN
GR
NI
SC
FR
IT
NI
BE
BE
TR
PT
IT
ES
GR
HU
WA
OK
MT
CH
CY
CY
FR
IT
WA
FR
BE
IT
CI
FR
FR
IT
IT
CH
EN
CH
BE
IT
CI
EN
CI
EN
CH
MT
CY
CH
DE
ITIT
IE
CI
CH
CI
MT
FR
IT
NL
MT
IE
WA
IM
IT
IT
IT
MT
FR
NL
CH
MT
AT
HU
EN
SC
CH
AT
CY
SC
FR
ITPT
HU
SC
BE
IT
MT
IT
BG
AT
CY
CY CY
GI
IT
CH
IT
RU
MT
RO
IT
EN
PT
SC
BE
TR
RU
IM
FR
WA
CH
IT
CI
IT
MK
CI
FR
BE
CI
CH
WA
GR
CY
PT
SC
CH
CH
HU
WA
BE
BE
IT
NL
NO
FR
CH
FR
GR
CI
IT
GI
GR
FR
PL
IT
CH
FR
IT
FR
CY
IT
GR
BE
IT
BE
CZ
EN
NL
IT
AT
FR
IT
MT
CH
CY
FR
IT
FR
IT
IT
CY
FR
CH
IT
BE
BE
IT
CH
CH
FR
MT
IT
GI
FR
FR
CH
DE
CH
IT
CH
IT
GR
IT
NL
SC
IT
GR
CY
GI
FR
FR
OK
MT
EN
MT
CZ
GR
FR
IT
IT
EN
IT
IT
SK
IT
MT
CI
IT
CH
FR
DK
AT
CH
BE
FR
IT
MT
GR
FR
IT
CY
FR
CH
BE
FR
IT
IT
MT
IT
BE
RS
MT
MT
GI
IT
FR
FR
IT
CH
FR
IT
FR
CH
FR
TR
MT
BE
IT
WA
FR
HU
FR
MT
FR
FR
CZ
CZ
IT
IT
FR
FR
IE
PT
GR
IT
FR
MT
SK
EN
GR
FR
CH
PT
FR
IT
MT
MT
MT
FR
IT
BE
CH
FR
MT
FR
FR
IT
IT
FR
MT
TR CY
IT
IT
CH
FR
MT
BE
MT
RO
ITCZ
CH
FR
FR
IT
RO
IT
BE
IT
MT
IT
DE
IT
HU
FR
AT
MT
AL
FR
IT
FR
HU
SC
FR
FR
ITPT
FR
FR
IT
ES
IT
IT
MT
GI
GI
MT
MT
FR
PT
IT
FR
CH
PT
MT
ES
IT
PT
MT
FR
IT
IT
IT
BE
MT
MT
MT
IT
MT
ITITCY
MT
FR
ES
MT
BEFR
IT
MT
IT
MT
ES
ES
PT
ES
ES
ES
ES
FR
GI
GI
IT
IT
MT
IT
IT
MT
ES
FR
FR
MT
ES
FR
ES
GI
GI
FR
MT
ES
ES
ITPT
FR
PT
GI
HR
ES
MT
GI
GI
PT
PT
IT
MT
ES
PT
ES
MT
IT
PT
MT
MT
MT IT
GI
MT
ES
FR
PT
FR
GI
ES
ES
PT
FR
PT
ES
PT
FR
ES
FR
PT
ES
FR
PT
ES
ES
ES
GI
ES
ES
ES
PT
FR
PT
ES
PT
GR
MT
PT
PT
PT
IT
ES
ES
PT
ES
PT
GI
PT
PT
ES
PT
ES
IT
FR
PT
PT
MT
MT
ES
PT
FR FR
PT
PT
ES
FR
ES
PT
ES
MT
ES
CH
ES
ES
ES
ES
PT
ES
ES
GI
PT
PT
PT
PT
PT
PT
GI
ES
PT
PT
PT
PT
PT
ES
FR
ES
GI
GI
ES
ES
ES
ES
ES
ES
ES
PT
GI
PT
GI
GI
GI
ES
PT
PT
FR
PT
ES
PT
ES
PT
PT
ES
PT
GI
PT
ES
PT
GI
PT
PT
PT
ES
PT
ES
ES
ES
ES
ES
PT
ES
GI
PT
PT
ES
ES
ES
ES
ES
GI
PT
ES
ES
PT
ES
ES
PT
GI
DE
PT
ES
PT
FR
ES
PT
ES
PT
ES
PT
GI
ES
ES
PT
PT
ES
ES
ES
PT
ES
PT
PT
ES
PT
ES
PT
ES
PT
ES
PT
PT
FR
ES
PT
PT
PT
ES
ES
ES
ES
PT
ES
PT
PT
PT
ES
ES
PT
ES
PT
ES
ES
GI
PT
PT
ESPT
GI
ES
ES
ES
PT
PT
PT
PT
FR
ES
ES
ESES
PT
ES
PT
ES
PT
ES
PT
PT
PT
ES
ES
ES
ES
ES
ES
PT
ES
ES
ES
ES
ES
PT
PT
ES
PT
PT
PT
FR
ES
PT
ES
ES
ES
PT
ES
ES
ES
PT
PTES
ES
CH
ES
ES
ES
ES
ES
ES PT
PT
ES
ES
ES
ESPT
ES
ESES
IT
FR
ES
UKB EUR
Group
a
Near East
a
SE Europe
a
S Europe
a
E Europe
a
C Europe
a
N Europe
a
W Europe
a
Brit. & Ire.
IT
IT
IT
IT
IT
−0.05
0.00
0.03
0.06
Principal Component 1
Supplementary Figure 2.11 – Comparison to 5,500 UK Biobank1 European individuals to 905 West
Eurasians from the Human Origins2 dataset with principal component analysis. UK Biobank individuals
were projected onto the genetic variation of the Human Origin references using PLINK4,5 (see
Methods). (A) The principal component (PC) coordinates of the Human origins reference individuals
for PC1 and PC2, with label shown by point colour and shape coding, and UK Biobank plotted as grey
points. (B) The PC coordinates of the UK Biobank individuals, colour coded according to European
meta-region and lettering to show country of birth.
126
127
128
129
130
131
132
Supplementary Data 3 – Comparison on Linked and Unlinked
Methods
Multiple authors have noted the increased power to detect fine-scale genetic structure in populations
when utilising “linked” methods (as discussed by Lawson et al7, Leslie et al8, and others using European
populations9-12). To elaborate, “linked” methods are methods which consider the linkage (in the
linkage disequilibrium sense) between markers and do not assume that genetic markers are
independent with respect to linkage disequilibrium (LD) – i.e., so-called “unlinked” methods.
133
134
135
136
137
138
139
140
141
142
143
144
Indeed, we observe a similar effect comparing the principal components of an unlinked PCA13 from
PLINK4,5 (Figure 1) and principal components calculated from a co-ancestry matrix from pbwt paint14
(Figure 2), our equivalent of a linked analysis. Genetic regions of Europe appear to separate out more
in this pbwt-based PCA than compared to the PCA calculated from unlinked allele-frequency data. As
mentioned in the main text Finland appears to be more strongly differentiated in PCA, possibly as a
result of the increased haplotype sharing within that population (see Figures 3 and 5). Other
populations such as Malta, Turkey and Cyprus, and differentiation between central mainland Europe
and Scandinavia to the north appear greater. We attribute this to PCA decomposition of the coancestry or genetic-relationship matrices detecting differing relationships along different time frames
as haplotype information would be expected to “up-weight” more recent relationships. To further
explore this we perform additional analysis, comparing unlinked and linked PCA methodologies on to
a well described dataset of European ancestry.
145
146
147
148
149
150
This dataset, the POPRES dataset15 was used previously by Novembre and colleagues16 using unlinked
methods to strikingly describe the overall European genetic landscape. We investigate whether our
results from pbwt-PCA are consistent with previous characterisations we analysed the POPRES
dataset15 utilised by Novembre et al16. We compared an unlinked method, performing PCA on a PLINK
generated genetic-relationship-matrix (GRM), to two linked methods - performing PCA on the coancestry chunkcounts matrices estimated by pbwt and CHROMOPAINTER7.
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
We selected individuals of European ancestry from versions 1 and 2 of the Population Reference
Sample (POPRES) dataset15 (n=5,917), a DNA resource from multiple studies across the world. We
filtered individuals based on the observed country of birth data of their grandparents, selecting
individuals whose grandparents were from the same European country. Due to the dataset containing
>1000 Swiss samples, we randomly down sampled this label, choosing 100 individuals who were SwissFrench.
168
169
170
We divided the unpruned dataset by chromosome and converted these subsets to VCF file format
using PLINK (ver. 1.9). These genotypes were then phased using SHAPEIT (ver. 4.2.1)18 with effective
population size set to 11,418 and using genome map build GRCh37. The phased genotype VCF files
Using PLINK (ver. 1.9)4,5, we excluded any strand-ambiguous SNPs (A/T, G/C) and kept autosomal SNPs.
Further, we removed SNPs with minor allele frequency (MAF) <2%, missingness >5%, and a HardyWeinberg Equilibrium (HWE) p-value <1e-6. Additionally, we removed individuals with SNP
missingness >5%. Pairs of related individuals (up to 3rd degree relations) were identified using KING
(ver. 2.2)17 and one random individual from the pair was removed. To calculate an unlinked PCA, we
utilised a set of SNPs pruned with respect to linkage disequilibrium using the PLINK command --indeppairwise 1000 50 0.2, performing PCA using the --pca PLINK command on these samples and pruned
markers (nsamples = 954, nmarkers= 103,925).
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
were converted to the hap/sample format using bcftools (ver. 1.4.1)19 which were subsequently
painted using (a), CHROMOPAINTER, which is a part of the fs (ver. 4.1.1) toolset7, and (b) pbwt paint
from the pbwt package14.
To generate the co-ancestry matrixes and perform PCA we carried out the following pipelines. For
CHROMOPAINTER, we converted the files hap/sample file set to the CHROMOPAINTER input format
using scripts provided by the authors of the fs utility. Next, the inferred mutation rate and effective
population size parameters of CHROMOPAINTER Hidden Markov Model were estimated on the
autosomal chromosomes (fs “stage 1”). Haplotype sharing in the full dataset was then estimated with
these parameters (fs “stage 2”) generating co-ancestry matrices of “chunk counts” and “chunk
lengths”, using default parameters. We calculated principal components on the co-ancestry matrix
generated by CHROMOPAINTER using R scripts provided by the authors of fs. Alternatively, for pbwt,
we used the phased VCF files to generate a co-ancestry matrix using the pbwt -paint command. We
calculated principal components on the chunkcounts output from the analysis using the same scripts
to perform the CHROMOPAINTER PCA, setting the diagonals of the pbwt paint co-ancestry matrix to
0. Additionally, we split the countries in the dataset into ten regional labels depending on their
geographic position within Europe (see below).
190
191
192
193
194
195
Supplementary Figure 3.1 - Principal component analysis13 of 954 Europeans selected from the
POPRES15 dataset, calculated from genetic relationship matrix output from PLINK4,5 (A), pbwt14 coancestry chunkcounts matrix (B), and CHROMOPAINTER7 co-ancestry chunkcounts matrix. Individual
genotypes are shown by letters which encode the alpha-2 ISO 3166 international standard codes, and
196
197
are colour coded according to geographic region. The median position for each country label is shown
as a label encapsulated by a circle.
198
199
200
201
202
203
204
205
206
207
In our analysis of comparing linked versus unlinked analyses, we show that European individuals do
indeed exhibit great overall separation in principal component space (panel A versus B or C in Figure
S3.1). All three PCAs show the defined north to south, and east to west gradients in European
genetics16,20,21, and agree with our own findings using an expanded sample of European genetics
(Figure 1). Our results from pbwt and CHROMOPAINTER show greater separation in south-east Europe,
particularly individuals with Kosovan or Yugoslavian grandparental birthplaces. As these analyses are
haplotype based this may be due to the relative enrichment of haplotype sharing in this region
compared to other populations sampled in the POPRES dataset, consistent with our findings of
individuals born from that European region (Figure 4).
208
209
210
211
Unfortunately, we were unable to analyse the genotypes of the exact individuals reported by
Novembre et al16, leading to a reduction of sample sizes notably from the UK. This may explain that
whilst we capture the general structure reported by Novembre et al, there are some topographic
differences compared to our Figure S3.1A.
212
213
214
215
216
217
218
219
220
221
222
223
224
Our comparative results show that each method captures the broad genetic structure in the same
dataset, though it appears exactly relationships are weighted differently, hence the differing
topologies in PC space. Ultimately, all are informative, though linked methods were able to separate
out sub-regions within the same dataset better than unlinked methods in both our analysis of the UK
Biobank1 and POPRES15 datasets. These results are in agreement with the initial report of the
CHROMOPAINTER method7, and findings from groups applying these methods to specific
populations8,22. This performance is thought as a reflection of haplotypes to better capture
information about more recent relatedness between individuals. Therefore, in this analysis of the
POPRES dataset, and the comparison between unlinked and linked methodologies, we are confident
both are sample capture the major genetic structure within Europe, and our findings from eigen
decomposition of the pbwt paint co-ancestry matrix is consistent with previous efforts in Europe16 and
findings from using the CHROMOPAINTER method7,8, which unfortunately doesn’t scale to this sample
size well.
225
226
227
228
229
230
231
232
233
234
235
236
Supplementary Data 4 - Genetic Structure of European Leiden
Clusters
Investigating the population structure captured in our sampled of UK Biobank1 participants with a
European place of birth, we performed Leiden3 clustering of a network of nodes made up of
individuals, and edges of the per-individual pair pbwt paint14 co-ancestry estimate (see Figure 2,
Methods). To further characterise these clusters, we perform several additional analyses, shown
below. The ADMIXTURE6 ancestry component estimates are the same calculated in Supplementary
Data 2.3, but individuals are instead grouped by Leiden cluster membership (Supplementary Figure
4.1). We relabelled the principal component coordinates for the 5,500 UK Biobank individuals
projected onto the genetic variation of west Eurasians from the Human Origins dataset2, labelling the
UK Biobank individuals by Leiden cluster (Supplementary Figure 4.2).
237
238
239
240
241
Supplementary Figure 4.1 - ADMIXTURE6 ancestry component estimates of 41 Leiden3 clusters of
5,500 UK Biobank1 participants over k values of two through to seven.
A
B
HO
Population
Jew_Turkish
Adygei
Jew_Yemenite
Balkar
Albanian
Chechen
Bulgarian
Georgian
Croatian
Iranian
Greek
Iranian_Bandari
Romanian
Kumyk
Basque
Lezgin
Italian_North
Nogai
Italian_South
Principal Component 2
North_Ossetian
0.00
−0.05
UKBB EUR
Cluster
0.05
Belgium & France
Channel Islands
Netherlands
CE Europe 1
Switzerland
CE Europe 2
Belgium
NE Balkans
Denmark
NW Balkans
Norway
Poland
Sweden 1
Russia
N.Ireland & Scotland 1
Latvia & Lithuania
N.Ireland & Scotland 2
Finland
N.Ireland & Scotland 3
Mixed Scand.
N.Ireland & Scotland 4
Italy
Orkney
Greece
Ireland 1
Albania & Greece
Ireland & N.Ireland
Turkey
Ireland 2
Cyprus
Isle of Man
Portugal
England 1
Spain
France
France & Switz.
England 2
Mixed European
Wales 1
Malta
Maltese
Armenian
Sardinian
Assyrian
Sicilian
Cypriot
Spanish
Druze
Spanish_North
Jordanian
English
Lebanese
French
Lebanese_Christian
Orcadian
Lebanese_Muslim
Scottish
Palestinian
Belarusian
Syrian
Czech
Turkish
Estonian
BedouinA
Hungarian
BedouinB
Ukrainian
Saudi
Chuvash
Yemeni
Finnish
Libyan
Icelandic
Moroccan
Lithuanian
Jew_Ashkenazi
Mordovian
Jew_Georgian
Norwegian
Jew_Iranian
Russian
Jew_iraqi
Saami_WGA
Principal Component 2
0.05
Abkhasian
0.00
−0.05
Wales 2
242
243
244
245
246
247
248
0.00
0.03
Principal Component 1
0.06
0.00
0.03
0.06
Principal Component 1
Supplementary Figure 4.2 – Principal component analysis of 5,500 UK Biobank participants with
European birth places, projecting onto West Eurasian references from the Human Origins dataset. (A)
Each coloured point represents the genotype of one Human Origins reference individual, with UK
Biobank individuals represented by grey points. (B). Each colour point represents the genotype of one
UK Biobank individual, with Human Origin references shown as grey points.
249
250
251
252
253
Supplementary Figure 4.3 - Heatmap of average per-cluster pair of haplotype copying profiles
between every Leiden cluster as a “target”, using every other Leiden cluster as a potential “source”
cluster. Average contribution was estimated using an adaptation of a nnls-based method8, using pbwt
paint co-ancestry estimates as copy vectors.
254
255
256
257
258
259
260
Supplementary Figure 4.4 - Plot of t-SNE23 decomposition of the top 10 principal components
calculated from the pbwt paint chunkcounts co-ancestry matrix. t-SNE analysis summarised haplotypic
relationships between 5,500 UK Biobank individuals with a European birthplace into two dimensions,
and individuals are colour and shape coded according to Leiden cluster membership.
261
262
263
264
Supplementary Data 5 – Detailed European Genetic Landscape
265
266
267
268
269
North-Western Europe
270
271
272
273
274
275
276
277
278
279
280
281
282
283
Individuals from Scotland and Northern Ireland are still grouped together at this level of clustering,
agreeing with previous observations of gene flow between Scotland and the north of Ireland24. This is
further supported in our nnls analysis (SI Appendix Fig. S4.3) which models the N.Ireland & Scotland
clusters with a contribution from the N.Ireland & Ireland cluster. We continue observations that
individuals with an Orcadian birthplace exhibit high haplotype sharing8,24 (Figure 3), and have a
historically low effective population size (Figure 4) – which also agrees with estimates and time of
lowest population size from an analysis of Orcadian individuals with extended recent ancestry from
those isles11. Interestingly, we identify a discrete cluster of individuals (n=41) predominantly from the
Isle of Man (Figure 2). These individuals present a similar profile of modest isolation as Orcadian
individuals, with a reduced population size from 30 to 10 generations (Figure 4), modest increase in
length of IBD segments (Figure 3), and elevated levels of ROH (Figure 5) - though the latter is not as
elevated as Orkney. Previous analysis of genotypes from the Isle of Man agrees with genetic
structure24, though we estimate elevated ROH in this sample compared to that separate of Manx
genotypes.
284
285
286
287
288
289
290
291
292
293
294
295
296
297
Clusters with English, Welsh, French and Channel Island membership are placed in the eastern British
Isles sub-branch. Wales, especially Wales 2, shows elevated within-cluster sharing of IBD-segment
count compared to other clusters on this sub-branch suggestive of modest isolation. This agrees with
previous analyses of samples with extended Welsh ancestry from the Peoples of the British Isles
Study8. As discussed in the main manuscript, individuals placed within the Channel Islands cluster
exhibit elevated sharing more of and longer IBD-segments than other Eng. & Wales clusters. These
IBD results reflect the sustained lower effective population size than clusters on the same sub-branch,
though the Channel Islands exhibit evidence of a more modest population contraction than Orkney.
The 117 individuals with a Channel Island birthplace are not just placed in this cluster of 83 individuals,
but also within English (n=11), Welsh (n=4), French (n=5), or Irish/Scottish (n=19) clusters. Compared
to Orkney, whose cluster shows equivalent IBD sharing to Channel Islands, individuals with a Channel
Islands birthplace are more distributed in other clusters – suggesting that whilst there may be an island
population of interest to haplotype mapping, the islands are less isolated which reflects their
geographic position in-between France and England.
298
299
300
301
302
303
304
Within the continental WE branch of clusters, Belgian, French, and German individuals are clustered
together (Belgium & France) and are grouped with clusters of separately Swiss (Switzerland) and Dutch
(Netherlands) predominant membership. A smaller cluster, Belgium (n=53), is detected. This cluster
does not appear to present high haplotype sharing or evidence of population contraction, and in nnls
analysis shares a haplotype profile predominantly donated from Netherlands. Further profiling of this
cluster is difficult, though there appears to be a genetic continuity between Belgium and Netherlands,
which is consistent with their geographic proximity and lack of large-scale geographic barriers.
For the purposes of brevity, in the main manuscript we highlight the primary results from each
geographic region sampled from the UK Biobank (UKBB). In this supplementary discussion we describe
in greater detail the results of each geographic region.
This group of clusters group individuals whose birthplace includes Scandinavia, the Low Countries,
France, Switzerland, and the British Isles and Ireland. Three main branches of clusters are detected,
one grouping Scandinavian individuals and other continental Europeans, one of individuals from the
eastern British Isles, and one of Scotland and Ireland.
305
306
307
Genetically, Swiss individuals project between France, Germany, and Italy in PCA, reflecting their
geographic location. We estimate Switzerland to have a recent effective population size to be
equivalent to Belgium or the Netherlands, but no evidence of substantial isolation.
308
309
310
311
312
313
314
315
316
317
318
319
320
321
Within the Scandinavian sub-branch, we detect three clusters which group individuals from Denmark,
Norway, and Sweden. These individuals project in-between northern continental Europe and Finnish
individuals, showing within-cluster IBD profiles equivalent to north-western Britain or Ireland. Our
small sample of Icelandic individuals (n=19) are grouped with Norwegians in Norway, and as described
in the main manuscript show evidence of isolation consistent with the population history of that
island25. Five out of six sampled Faroese individuals are grouped into the Sweden cluster. Similarly to
the Icelandic individuals in Norway, these Faroe Islanders show substantially higher IBD-segment
sharing than Swedish individuals placed in the same cluster. The average pair of Faroe-Faroe
individuals share 90 cM of IBD segments > 1cM, sharing 28 segments each of which are on average
3.2 cM long. The average Swedish-Swedish individual pair share a total of 21 cM over 14 segments,
each of which are 1.5 cM long. The low sample size of the Faroe Islands means it is unclear just how
representative these results are, though a recent study Faroese genotypes indicates the population
does indeed exhibit unsurprising hallmarks of isolation26 – though we are unable to extend and refine
this picture as has been possible in Malta.
322
323
324
325
326
327
Central-Eastern Europe
328
329
330
331
332
333
334
335
336
337
338
339
Within the CE European group of clusters there are two sub-groups, each with two clusters. The first,
CE Europe 1 and 2 cluster individuals predominantly from Germany, Austria, Poland, the Czech
Republic, and Hungary. CE Europe 1 appears to group individuals of a more easterly birthplace, and in
nnls analysis its haplotype profile includes more of a contribution from the Poland cluster, whereas CE
Europe 2 contains more individuals from Austria and Germany and a slightly higher contribution from
Switzerland in nnls analysis. In analysis of within-cluster haplotype sharing CE Europe 1 shows modest
elevation of IBD-segment sharing (Figure 3) consistent with a slightly lower historical population size,
which our Ne estimates from 30 generations ago agree with. Within the north of the Balkans, we
observe a general east-west divide in our clustering and PCA results (Figure 2). With IBD-segment
sharing as well as ROH information, it appears NW Balkans shows modestly elevated levels of
haplotype sharing consistent with a lower effective population size, this is supported by a consistently
lower effective population size in NW Balkans compared to NE Balkans.
340
341
342
343
344
345
346
347
348
349
In our main manuscript we discuss the results of the Finnish, E Europe, and northern Balkans clusters
in greater detail, noting that our results in Finland agree with existing literature on this well studied
population. In NE Europe the UK Biobank sample largely agrees with the Human Origins references,
though our Russian samples more project towards Baltic/Ukrainian genetic space as opposed to the
Human Origins Russians, who project more towards Chuvash or Saami – though there is overlap in the
UKBB and Human Origins Russians. We suspect that the UKBB sample of Russian ancestry is therefore
biased towards European ancestry rather than the central Eurasian or Caucus/related ancestry that is
represented by the Chuvash or Saami in the Human Origins West Eurasian sample. This heterogeneity
in sampled Russian ancestry can be explained by the variation in sampled ancestries within the Russian
Federation, with ancestry in samples closer to Europe exhibiting more European-modal ancestry27.
The second branch of Leiden clusters groups individuals with a birthplace from the centre and east of
Europe, from Germany in the west to Russia in the east. This branch contains several sub-branches
which group individuals from geographically adjacent locations in Europe; NE Europe (with Baltic,
Polish, and Russian membership), CE Europe (with membership from the north of the Balkans and the
centre and east of Europe), and Finland as an outgroup.
350
351
352
353
354
355
356
357
358
359
Southern Europe
360
361
362
363
364
365
Geographically neighbouring Malta, individuals from Italy are grouped into one Leiden cluster, though
PCA divides individuals into predominantly two clusters, which appear to reflect the general northsouth divide previously detected in Italian genetics10. Italy haplotype contributions come from the
mixed France & Switz. cluster, as well as Turkey and Greece - which agrees with its position between
western Europe and the south-east of Europe and the Near East10. This is associated with a consistently
high historical effective population size compared to other European regions (Figure 4).
366
367
368
369
370
371
372
373
374
375
376
Within Iberia, the predominant divide is between Spanish and Portuguese samples, in agreement with
recent analyses12. Spain and Portugal show evidence of reciprocal admixture in nnls analysis. Portugal
appears to have modestly elevated haplotype sharing compared to Spain, consistent with its
geographic position at the end of the Iberian Peninsula. This haplotype sharing is observed in both IBD
and ROH segment sharing, where elevation of the latter appears to be driven by predominantly longer
ROH than greater numbers of ROH (Figure 5). As discussed in the main manuscript we detect small
group of Spanish individuals within Spain which is genetically distinct from other Spanish clusters and
exhibits elevated IBD-segment sharing. With evidence from co-analysis of references from the Human
Origins dataset, we suspect that this cluster represents ancestry from the north-east of Spain and the
Basque population. Further sample annotation is not available for these samples, so we are unable to
definitively prove this.
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
Within SE Europe, we cluster individuals from Albania, Greece, North Macedonia, Turkey. Italian
clusters from an outgroup to this sub-branch along with Cyprus. We observe some signal of haplotype
sharing between Italy and Greece, as well as sharing between Italy and Turkey. As discussed in the
main manuscript, we detect a small cluster of Greek, Albanian, and North Macedonian membership,
Albania & Greece. It is unclear if this small cluster is largely represented of this geographic region,
though there is no obvious reason that the UKBB would have sampled a specific community from this
area of Europe with history of isolation. Turkey and Cyprus exhibit evidence of isolation that is
consistent to some degree of consanguinity in inbreeding coefficient analysis. In comparison to Turkish
references from the Human Origins dataset Turkey co-segregates with other Turkish references –
forming a tight cluster in PCA in the centre of the distribution of Turkish genotypes. We see a similar
relationship between Cyprus and Cypriot Human Origins references. This appears to be a relatively
isolated Turkish community, captured within the UK Biobank. Lastly, as further expanded upon in the
main manuscript as well as in SI Appendix Data 7 we detect a connected community of individuals
with a mixture of birthplaces across Europe. In analysis of ancestry, population structure, and
characterisation of demographic history we propose this Mixed European cluster represents a
community of Ashkenazi Jewish participants in the UK Biobank. For more information and analysis,
see SI Appendix Data 7.
394
The third and final group of Leiden clusters contains individuals with a birth-place from regions within
or north of the Mediterranean. The major sub-groups include the islands of Malta and Cyprus
(grouped for convenience, but genetically distinct), Italy, the Iberian Peninsula, SE Europe – which
includes Greece and Turkey, and a cluster of individuals with a heterogenous mixture of birthplaces
(Mixed European). We discuss the full results of the large Malta cluster in our main manuscript – but
briefly our results show that Malta is genetically distinct and presents evidence of genetic isolation
with high haplotype sharing and genetic distances to non-Maltese clusters. In PCA with UKBB
individuals, or projected onto Human Origins genetic variation, Maltese individuals project close with
Cyprus 2 and Italy 2.
395
396
397
398
399
400
401
402
Supplementary Data 6 - Malta
403
404
405
406
407
408
409
Investigating the population structure in this sample of European genotypes we observe population
structure within our Maltese sample (see Figure 1, and Figure S2.2), where three broad clusters
emerge in principal component space. Further investigating this, we find that principal component
(PC) seven of the PLINK4,5 PCA, of which PC one and two are shown in Figure 1, also separates these
individuals, see Figure S2.2. When we code the sampled Maltese individuals instead by Leiden cluster
membership of the Malta cluster or not, the Leiden clustering clearly separates out these two groups,
see below:
In analysis of European ancestry sampled within the UK Biobank, we report to our knowledge the
largest dataset of Maltese genotypes publicly available. As part of the sampling scheme, we randomly
down-sampled individuals from countries/regions of birth which had more than 200 individuals to a
randomly selected 200 individuals. This was to both reduce sample size and to control for uneven
sampling rates across European regions, for example there were over 1000 individuals with a German
place-of-birth compared to 180 Swiss. We randomly sampled 200 Maltese individuals out of a possible
362.
410
411
412
413
414
415
416
417
418
Where the Malta cluster is comprised of the Maltese individuals at the terminus of the cline away
from other Europeans, as well as the smaller cluster of Maltese individuals which project in-between
that group of Maltese and other Europeans. These two groups of Maltese individuals within the Malta
Leiden cluster also show two patterns of IBD sharing within the cluster (Figure 3), with the
intermediate Maltese individuals in PC seven also intermediate in within-cluster IBD-sharing
estimates.
419
420
421
422
423
424
425
426
427
428
429
430
Supplementary Data 7 - Mixed Europeans
431
432
433
434
435
436
437
438
In addition to this genetic structure, individuals within this cluster exhibit elevated within-cluster IBDsegment sharing (Figure 3), a historically low effective population size with recent population
expansion (Figure 4) and elevation of ROH sharing. This is consistent with a small isolate population
that has experienced a population size bottleneck around 30 generations ago. These demographic
results, with analysis of population structure in the context of Europe and the Middle East is suggestive
that this is a community of Ashkenazi Jews. This would be consistent with previous analyses of the UK
Biobank which detected such a community28, as well as the population history of European Ashkenazi
Jews as previous modelled in more detail with individuals recruited from that community29.
439
440
441
442
443
444
445
446
447
448
To further confirm this, we performed an additional analysis with references from the Human Origins
dataset. Using the same methodology to project the UK Biobank individuals onto western Eurasian
genetic variation, we selected Human Origin West Eurasian individuals, as well as Yoruban references
form that dataset, and merged this genotype data with the UK Biobank dataset, leaving 6,476
individual genotypes of 61,217 common SNPs after the same quality control thresholds as in the
projected PC analysis. Analysing a subset of these markers which were in approximate linkage (filtering
SNPs with the PLINK4,5 option --indep-pairwise 1000 50 0.2) we tested allelic sharing using f-statistics
implemented in the R package ADMIXTOOLS2 (manuscript in prep). For each Leiden cluster we tested
allelic sharing between the closest population reference in the Human Origins dataset and Human
Origin Ashkenazi Jew references in the form:
449
f4(Yoruban, Y; Ashkenazi Jew, European Reference)
450
451
452
453
454
455
456
457
The f4 estimates for each Leiden cluster are shown below. Along the y-axis is each Leiden cluster tested
in “quotation marks” and the paired population reference chosen from the Human Origins dataset.
“Mixed A” and “Mixed B” refer to two sub-groups of the Mixed European cluster which are expanded
upon below. With the exception of Malta, Turkey, and Cyprus (i.e., the UK Biobank clusters closest to
Near Eastern populations in PCA) all European clusters are significantly differentiated from the Human
Origin Ashkenazi Jew references. Both Mixed European groups, Mixed A and Mixed B shared
significantly more alleles with Ashkenazi Jew than Turkish references as well as sharing more allelic
drift with Ashkenazi Jew references than Turkish Jews (see below).
In our analysis of European genotypes sampled in the UK Biobank1 we have identified a genetically
connected community of individuals with birthplaces across Europe, though with a slight bias towards
central/eastern European countries of birth. Despite being geographically dispersed, as represented
by birthplace information, these individuals project along a single principal component (principal
component nine in Figure S2.9) as well as are grouped as a single cluster by application of the Leiden
algorithm (Figure 2). Compared to the rest of the European sample, a core group of these Mixed
European individuals project towards individuals with a Near East birthplace (Figure 2), and project
towards Ashkenazi or Turkish Jewish references from the Human Origins2 dataset. Furthermore, in our
nnls based analysis (see Methods, Figure S4.3), the haplotype sharing of these individuals can be
modelled as a mixture of Near Eastern (Turkey: 0.12), southern European, (Italy: 0.10, Spain: 0.10),
and central or Eastern European sources (Poland: 0.14, CE European 1: 0.13, Russia: 0.07).
458
459
460
461
462
463
464
In PCA of the pbwt paint co-ancestry matrix, we observed this Mixed European cluster separate along
PC three, see below. Similarly to individuals placed in the Malta cluster, there appears to be two
groups of individuals along this PC, which we divided into two groups, Mixed A (Mixed European
individuals to the left of the red line below) and Mixed B (Mixed European individuals to the right of
the red line).
465
466
467
468
469
470
471
We next tested if these two groups within Mixed European presented different haplotype sharing
profiles with the rest of our European UK Biobank dataset. We performed a modification of our nnls
analysis (see Methods), this time modelling each Mixed A and Mixed B cluster as a mixture of any
other Leiden cluster (with the exception of the other Mixed European sub-group). We present the
results below, finding that the Mixed B cluster presents a more heterogenous sharing profile
consistent with admixture (which would be consistent with ADMIXTURE estimates (see SI Appendix
Data 4)).
472
473
474
475
476
477
478
Further testing evidence of Mixed A grouping individuals more genetic isolated we recorded levels
Runs of Homozygosity (ROH) between the groups, finding that on average individuals placed in Mixed
A carry more, and longer, ROH than individuals. These differences are significant comparing either the
average length of ROH an individual carries between the groups (Mann-Whitney U p-value: <0.0001)
or the number of ROH (Mann-Whitney U p-value: <0.0001).
479
480
481
482
483
484
485
486
We further explored the degree of haplotype sharing between and within these groups by recording
the total length and number of IBD-segments > 3 cM and < 30 cM29 shared within and between Mixed
A and Mixed B and a comparative cluster, England 2 using the same methodology as shown in Figure
3 and described in Methods. We find that IBD segments are slightly longer within and between Mixed
A and Mixed B (below), and that the average Mixed A pair of individuals (i.e., individuals both placed
in the Mixed A cluster) share a number higher number of IBD segments than the average pair of Mixed
B individuals. Interestingly the average Mixed A - Mixed B pair of individuals share more IBD segments
than the average Mixed B pair, suggesting the Mixed B are heterogenous within that group.
487
488
489
490
491
492
Lastly, we estimated the effective population size of each of the clusters Mixed A and Mixed B using
IBD-segments and IBDNe30, characterising the haplotype sharing and population structure in terms of
historical population size. We find that whilst both present evidence of historical population
contraction, we record a much minimal historical population size for Mixed A and Mixed B, consistent
with greater isolation.
493
494
495
Together these results suggest these two groups represent structure within the sampled community
from the UK Biobank, one more isolated than the other. The latter which has experience greater
496
497
498
admixture with European regions, and whose members are less genetically connected (as measure by
IBD segment sharing) than the former.
499
Supplementary Data 8 - IBD Ne Curves
500
501
502
503
504
Supplementary Figure 8.1 – Historical effective population size (Ne) of Leiden clusters grouping
individuals with a north-western European birthplace. Each panel shows the estimate for one cluster,
with grey curves showing the estimates for all other north-west European clusters. Shading indicates
the 95% confidence intervals for the estimates.
505
506
507
508
509
510
Supplementary Figure 8.2 – Historical effective population size (Ne) of Leiden clusters grouping
individuals with a northern Britain or Irish birthplace. Each panel shows the estimate for one cluster,
with grey curves showing the estimates for all other northern British or Irish clusters. Shading indicates
the 95% confidence intervals for the estimates.
511
512
513
514
515
516
Supplementary Figure 8.3 – Historical effective population size (Ne) of Leiden clusters grouping
individuals with an English or Welsh birthplace. Each panel shows the estimate for one cluster, with
grey curves showing the estimates for all other southern British clusters. Shading indicates the 95%
confidence intervals for the estimates.
517
518
519
520
521
522
Supplementary Figure 8.4 – Historical effective population size (Ne) of Leiden clusters grouping
individuals with a central, northern, or eastern European birthplace. Each panel shows the estimate
for one cluster, with grey curves showing the estimates for all other central, northern, or eastern
European clusters. Shading indicates the 95% confidence intervals for the estimates.
523
524
525
526
527
528
529
Supplementary Figure 8.5 – Historical effective population size (Ne) of Leiden clusters grouping
individuals with a southern European birthplace. Each panel shows the estimate for one cluster, with
grey curves showing the estimates for all other south European clusters. Shading indicates the 95%
confidence intervals for the estimates.
530
531
532
533
534
535
536
537
Supplementary Data 9 - Spain
In analysis of within-cluster Identity by Descent (IBD) segment analysis we detect a number of
individuals placed in the Spain cluster who present high levels of within-cluster IBD sharing compared
to other individuals with a birthplace in Spain (Figure 3). In principal component analysis (PCA) of the
pbwt paint co-ancestry matrix we observe a group of 12 Spain individuals who project away from the
rest of the cluster on principal component (PC) one, as well as PC six. This is shown below, with the
fifth and sixth PCs plotted and individuals labelled by Leiden cluster, and a horizontal black line
indicating the cut-off threshold for identifying “Spain Outliers” along PC six.
538
539
540
541
542
When we plot PCs one and two of the PCA decomposition of the pbwt paint co-ancestry matrix,
highlighting these 12 individuals, we find that these individuals are the Spain outliers along PC one as
well, see below. The Spain outliers are highlighted in black.
543
544
545
546
547
548
Due to the high levels of IBD segment sharing amongst these 12 individuals (Figure 3), we recorded
the levels of IBD segment sharing between these Spain outlier individuals and the rest of the Spain
cluster, as well as the levels of sharing between the two Spanish groups. We recorded total length of
IBD (below left), and the number of IBD segments (below right), finding that the Spain outliers indeed
share on average more IBD segments within, as well as between the outliers and the general Spain
cluster, than the average within Spain pair of individuals.
549
550
551
This increase in haplotype sharing is also mirrored in levels of Runs of Homozygosity (ROH), see below,
where the average total length of ROH > 1.5 Mb in length is elevated in the Spain outliers.
552
553
554
555
Furthermore, when projected onto the genetic variation of west Eurasian population references from
the Human Origins dataset2 the Spain outlier individuals project towards the variation of Basque and
northern Spanish references, see below.
556
557
558
559
560
561
562
563
Given the genetic distinctiveness of these outliers, their elevated haplotype sharing consistent with a
degree of isolation relative to the other UK Biobank individuals with a Spanish place of birth, and their
projection towards Basque genetic space in PCA, we infer these individuals to be of northern Spanish
or Basque ancestry. Previous genetic analyse of the Iberian Peninsula12 has shown this area of Spain
to be genetically distinct and would be consistent with the profile presented by these twelve
individuals sampled from the UK Biobank.
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature
562, 203-209 (2018).
Lazaridis, I. et al. Genomic insights into the origin of farming in the ancient Near East. Nature
536, 419-24 (2016).
Traag, V.A., Waltman, L. & van Eck, N.J. From Louvain to Leiden: guaranteeing wellconnected communities. Sci Rep 9, 5233 (2019).
Purcell, S. et al. PLINK: a tool set for whole-genome association and population-based
linkage analyses. Am J Hum Genet 81, 559-75 (2007).
Chang, C.C. et al. Second-generation PLINK: rising to the challenge of larger and richer
datasets. Gigascience 4, 7 (2015).
Alexander, D.H., Novembre, J. & Lange, K. Fast model-based estimation of ancestry in
unrelated individuals. Genome Res 19, 1655-64 (2009).
Lawson, D.J., Hellenthal, G., Myers, S. & Falush, D. Inference of population structure using
dense haplotype data. PLoS Genet 8, e1002453 (2012).
Leslie, S. et al. The fine-scale genetic structure of the British population. Nature 519, 309-314
(2015).
Karakachoff, M. et al. Fine-scale human genetic structure in Western France. Eur J Hum
Genet 23, 831-6 (2015).
Raveane, A. et al. Population structure of modern-day Italians reveals patterns of ancient
and archaic ancestries in Southern Europe. Sci Adv 5, eaaw3492 (2019).
Pankratov, V. et al. Differences in local population history at the finest level: the case of the
Estonian population. Eur J Hum Genet 28, 1580-1591 (2020).
Bycroft, C. et al. Patterns of genetic differentiation and the footprints of historical migrations
in the Iberian Peninsula. Nat Commun 10, 551 (2019).
Patterson, N., Price, A.L. & Reich, D. Population structure and eigenanalysis. PLoS Genet 2,
e190 (2006).
Durbin, R. Efficient haplotype matching and storage using the positional Burrows-Wheeler
transform (PBWT). Bioinformatics 30, 1266-72 (2014).
Nelson, M.R. et al. The Population Reference Sample, POPRES: a resource for population,
disease, and pharmacological genetics research. Am J Hum Genet 83, 347-58 (2008).
Novembre, J. et al. Genes mirror geography within Europe. Nature 456, 98-101 (2008).
Manichaikul, A. et al. Robust relationship inference in genome-wide association studies.
Bioinformatics 26, 2867-73 (2010).
Delaneau, O., Zagury, J.F., Robinson, M.R., Marchini, J.L. & Dermitzakis, E.T. Accurate,
scalable and integrative haplotype estimation. Nat Commun 10, 5436 (2019).
Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10(2021).
Lazaridis, I. et al. Ancient human genomes suggest three ancestral populations for presentday Europeans. Nature 513, 409-13 (2014).
Haak, W. et al. Massive migration from the steppe was a source for Indo-European
languages in Europe. Nature 522, 207-11 (2015).
Byrne, R.P. et al. Insular Celtic population structure and genomic footprints of migration.
PLoS Genet 14, e1007152 (2018).
van der Maaten, L.J.P. & Hinton, G.E. Visualizing high-dimensional data using t-SNE. J. Mach.
Learn. Res 9, 2579–605 (2008).
Gilbert, E. et al. The genetic landscape of Scotland and the Isles. Proc Natl Acad Sci U S A
116, 19064-19070 (2019).
Ebenesersdottir, S.S. et al. Ancient genomes from Iceland reveal the making of a human
population. Science 360, 1028-1032 (2018).
Leblond, C.S. et al. Both rare and common genetic variants contribute to autism in the Faroe
Islands. NPJ Genom Med 4, 1 (2019).
615
616
617
618
619
620
621
622
623
27.
28.
29.
30.
Khrunin, A.V. et al. A genome-wide analysis of populations from European Russia reveals a
new pole of genetic diversity in northern Europe. PLoS One 8, e58552 (2013).
Naseri, A. et al. Personalized genealogical history of UK individuals inferred from biobankscale IBD segments. BMC Biol 19, 32 (2021).
Xue, J., Lencz, T., Darvasi, A., Pe'er, I. & Carmi, S. The time and place of European admixture
in Ashkenazi Jewish history. PLoS Genet 13, e1006644 (2017).
Browning, S.R. & Browning, B.L. Accurate Non-parametric Estimation of Recent Effective
Population Size from Segments of Identity by Descent. Am J Hum Genet 97, 404-18 (2015).