Explore Your DNA

[kocovic]

Evo, napravio sam alat koji vam odredjuje kosinusnu sličnost:
https://v1rtualalloc.github.io/genetic-similarity/

[Небо_Сав]

Evo optimizovane liste da smanje šum:

1) Severozapadna Evropa: Celtic ili prosek Germanic/Gaulish. Ovo pokriva Kelto-Germansko jezgro (visok udeo stepe i rani evropski zemljoradnici).
2) Baltoslovenska grupa: Baltic jer ima najčistiji signal za severoistočnu Evropu. Baltic je ovde sidro za slovensku genetiku.
3) Italo-Iberijci: Italic ili Etruscan jer predstavlja mediteranski profil. Ovde bi se matematički uklopio i Paleo-Balkanic zbog sličnosti, tako da sa Balkana uzimamo Paleo-Balkan.
4) Istočna stepa: Avar, jer je ključan za detekciju istočno-evroazijskog uticaja (avarski period).
5) Uralic: Sadrži specifičan sibirsko-uralski pomak, različit od standardnog slovenskog signala.
6) Južni Mediteran: Phoenician i North-African. Detektuje afro-azijsku i levantsku komponentu koja se razlikuje od evropske.

Tako da bi konačni kalkulator mogao da bude nešto ovako:

Код: [Изабери]

Celtic,0.127028,0.134841,0.060663,0.048063,0.037119,0.017610,0.004106,0.003811,0.005293,0.004777,-0.005678,0.005541,-0.013977,-0.011765,0.019752,0.002326,-0.008887,0.002277,0.003567,0.002808,0.004824,0.002717,-0.001860,0.004879,-0.001028
Baltic,0.1324142,0.1252487,0.0944687,0.0996457,0.051189,0.0390447,0.0144533,0.0183068,-0.0025225,-0.0410637,-0.0027605,-0.0173595,0.0349107,0.0378692,-0.0137983,0.0003978,0.0041507,0.000169,0.0015293,0.0035227,-0.0033898,-0.0078518,0.0082372,-0.0142187,-0.0010177
Avar,0.042704,-0.374514,0.091613,-0.021495,-0.080668,-0.046419,0.017259,0.024745,0.008927,0.011123,-0.011503,0.000701,-0.004218,0.003188,0.002604,0.001067,0.000382,-0.003404,0.001881,0.011153,-0.023158,-0.005006,-0.019134,-0.002583,0.001635
Uralic,0.1013025,-0.0766725,0.1170643,0.0798887,-0.0298259,0.0001861,0.0050721,0.0140959,0.0024544,-0.031132,0.0352926,-0.0101536,0.0188426,-0.0285223,-0.0088781,0.0006629,-0.0022493,0.0013407,-0.0062852,-0.0061384,0.0164397,0.0022464,-0.001366,0.0015061,0.0013972
Phoenician,0.0896355,0.1544874,-0.0500626,-0.093024,-0.011579,-0.036186,-0.004377,-0.0089709,0.0143678,0.0110252,0.0088298,-0.0055264,0.0149775,0.0034232,-0.0122656,0.0037456,-0.0024936,0.0012989,0.0037552,-0.0014069,-0.0009982,0.0043125,-0.0051301,0.0010091,-0.0021105
Paleo-Balkanic,0.123008,0.147855,0.023861,-0.015954,0.030195,-0.007191,0.001060,0.000495,0.006273,0.027387,0.000031,0.007595,-0.013702,-0.003935,-0.008290,-0.003921,0.007116,0.001854,0.006836,-0.005574,-0.007092,0.004808,-0.000243,0.004937,-0.004113
North-African,-0.072847,0.142174,0.003017,-0.072352,0.034776,-0.037371,-0.034312,0.008077,0.08999,0.032985,0.011367,-0.010191,0.027354,-0.016515,0.024701,-0.01485,0.001825,-0.026605,-0.053799,0.008004,-0.015847,-0.044515,0.020829,-0.012532,0.01449


Uradio sam proračune za sve pripadnike roda Ozrinića, na osnovu G25 koordinata Davidskog sa ovim tvojim koordinatama, i uporedio sa ranijim rezultatima, kao i sa Illustrative DNA.

Ove koordinate favorizuju Paleo-Balkanic komponentu, i to tako što dodaju od 5-14% Peleo-Palkanske komponente na uštrb Eearly Slavic (odnosno njenu zamenu - Baltic).
To jest, kad nema Early Slavic - raste Paleo-Balkanic udeo, tj ovaj Baltic ne pokriva dobro Eearly Slavic komponentu.
 Rekao bih da je šum povećan, a ne smanjen.

[kocovic]

Uradio sam proračune za sve pripadnike roda Ozrinića, na osnovu G25 koordinata Davidskog sa ovim tvojim koordinatama, i uporedio sa ranijim rezultatima, kao i sa Illustrative DNA.

Ove koordinate favorizuju Paleo-Balkanic komponentu, i to tako što dodaju od 5-14% Peleo-Palkanske komponente na uštrb Eearly Slavic (odnosno njenu zamenu - Baltic).
To jest, kad nema Early Slavic - raste Paleo-Balkanic udeo, tj ovaj Baltic ne pokriva dobro Eearly Slavic komponentu.
 Rekao bih da je šum povećan, a ne smanjen.

Da, to je normalno, jer su rani Sloveni 65-70% Baltic + 30-35% Paleo-Balkan.

[Небојша]

Evo optimizovane liste da smanje šum:

1) Severozapadna Evropa: Celtic ili prosek Germanic/Gaulish. Ovo pokriva Kelto-Germansko jezgro (visok udeo stepe i rani evropski zemljoradnici).
2) Baltoslovenska grupa: Baltic jer ima najčistiji signal za severoistočnu Evropu. Baltic je ovde sidro za slovensku genetiku.
3) Italo-Iberijci: Italic ili Etruscan jer predstavlja mediteranski profil. Ovde bi se matematički uklopio i Paleo-Balkanic zbog sličnosti, tako da sa Balkana uzimamo Paleo-Balkan.
4) Istočna stepa: Avar, jer je ključan za detekciju istočno-evroazijskog uticaja (avarski period).
5) Uralic: Sadrži specifičan sibirsko-uralski pomak, različit od standardnog slovenskog signala.
6) Južni Mediteran: Phoenician i North-African. Detektuje afro-azijsku i levantsku komponentu koja se razlikuje od evropske.

Tako da bi konačni kalkulator mogao da bude nešto ovako:

Код: [Изабери]

Celtic,0.127028,0.134841,0.060663,0.048063,0.037119,0.017610,0.004106,0.003811,0.005293,0.004777,-0.005678,0.005541,-0.013977,-0.011765,0.019752,0.002326,-0.008887,0.002277,0.003567,0.002808,0.004824,0.002717,-0.001860,0.004879,-0.001028
Baltic,0.1324142,0.1252487,0.0944687,0.0996457,0.051189,0.0390447,0.0144533,0.0183068,-0.0025225,-0.0410637,-0.0027605,-0.0173595,0.0349107,0.0378692,-0.0137983,0.0003978,0.0041507,0.000169,0.0015293,0.0035227,-0.0033898,-0.0078518,0.0082372,-0.0142187,-0.0010177
Avar,0.042704,-0.374514,0.091613,-0.021495,-0.080668,-0.046419,0.017259,0.024745,0.008927,0.011123,-0.011503,0.000701,-0.004218,0.003188,0.002604,0.001067,0.000382,-0.003404,0.001881,0.011153,-0.023158,-0.005006,-0.019134,-0.002583,0.001635
Uralic,0.1013025,-0.0766725,0.1170643,0.0798887,-0.0298259,0.0001861,0.0050721,0.0140959,0.0024544,-0.031132,0.0352926,-0.0101536,0.0188426,-0.0285223,-0.0088781,0.0006629,-0.0022493,0.0013407,-0.0062852,-0.0061384,0.0164397,0.0022464,-0.001366,0.0015061,0.0013972
Phoenician,0.0896355,0.1544874,-0.0500626,-0.093024,-0.011579,-0.036186,-0.004377,-0.0089709,0.0143678,0.0110252,0.0088298,-0.0055264,0.0149775,0.0034232,-0.0122656,0.0037456,-0.0024936,0.0012989,0.0037552,-0.0014069,-0.0009982,0.0043125,-0.0051301,0.0010091,-0.0021105
Paleo-Balkanic,0.123008,0.147855,0.023861,-0.015954,0.030195,-0.007191,0.001060,0.000495,0.006273,0.027387,0.000031,0.007595,-0.013702,-0.003935,-0.008290,-0.003921,0.007116,0.001854,0.006836,-0.005574,-0.007092,0.004808,-0.000243,0.004937,-0.004113
North-African,-0.072847,0.142174,0.003017,-0.072352,0.034776,-0.037371,-0.034312,0.008077,0.08999,0.032985,0.011367,-0.010191,0.027354,-0.016515,0.024701,-0.01485,0.001825,-0.026605,-0.053799,0.008004,-0.015847,-0.044515,0.020829,-0.012532,0.01449


Target: Небојша_SK
Distance: 2.0274% / 0.02027352
54.0   Paleo-Balkanic
32.2   Baltic
12.4   Phoenician
1.4   Celtic

Target: Мајка
Distance: 3.0175% / 0.03017471
56.6   Paleo-Balkanic
30.6   Baltic
12.8   Phoenician

[Stefan Dragićević]

Evo optimizovane liste da smanje šum:

1) Severozapadna Evropa: Celtic ili prosek Germanic/Gaulish. Ovo pokriva Kelto-Germansko jezgro (visok udeo stepe i rani evropski zemljoradnici).
2) Baltoslovenska grupa: Baltic jer ima najčistiji signal za severoistočnu Evropu. Baltic je ovde sidro za slovensku genetiku.
3) Italo-Iberijci: Italic ili Etruscan jer predstavlja mediteranski profil. Ovde bi se matematički uklopio i Paleo-Balkanic zbog sličnosti, tako da sa Balkana uzimamo Paleo-Balkan.
4) Istočna stepa: Avar, jer je ključan za detekciju istočno-evroazijskog uticaja (avarski period).
5) Uralic: Sadrži specifičan sibirsko-uralski pomak, različit od standardnog slovenskog signala.
6) Južni Mediteran: Phoenician i North-African. Detektuje afro-azijsku i levantsku komponentu koja se razlikuje od evropske.

Tako da bi konačni kalkulator mogao da bude nešto ovako:

Код: [Изабери]

Celtic,0.127028,0.134841,0.060663,0.048063,0.037119,0.017610,0.004106,0.003811,0.005293,0.004777,-0.005678,0.005541,-0.013977,-0.011765,0.019752,0.002326,-0.008887,0.002277,0.003567,0.002808,0.004824,0.002717,-0.001860,0.004879,-0.001028
Baltic,0.1324142,0.1252487,0.0944687,0.0996457,0.051189,0.0390447,0.0144533,0.0183068,-0.0025225,-0.0410637,-0.0027605,-0.0173595,0.0349107,0.0378692,-0.0137983,0.0003978,0.0041507,0.000169,0.0015293,0.0035227,-0.0033898,-0.0078518,0.0082372,-0.0142187,-0.0010177
Avar,0.042704,-0.374514,0.091613,-0.021495,-0.080668,-0.046419,0.017259,0.024745,0.008927,0.011123,-0.011503,0.000701,-0.004218,0.003188,0.002604,0.001067,0.000382,-0.003404,0.001881,0.011153,-0.023158,-0.005006,-0.019134,-0.002583,0.001635
Uralic,0.1013025,-0.0766725,0.1170643,0.0798887,-0.0298259,0.0001861,0.0050721,0.0140959,0.0024544,-0.031132,0.0352926,-0.0101536,0.0188426,-0.0285223,-0.0088781,0.0006629,-0.0022493,0.0013407,-0.0062852,-0.0061384,0.0164397,0.0022464,-0.001366,0.0015061,0.0013972
Phoenician,0.0896355,0.1544874,-0.0500626,-0.093024,-0.011579,-0.036186,-0.004377,-0.0089709,0.0143678,0.0110252,0.0088298,-0.0055264,0.0149775,0.0034232,-0.0122656,0.0037456,-0.0024936,0.0012989,0.0037552,-0.0014069,-0.0009982,0.0043125,-0.0051301,0.0010091,-0.0021105
Paleo-Balkanic,0.123008,0.147855,0.023861,-0.015954,0.030195,-0.007191,0.001060,0.000495,0.006273,0.027387,0.000031,0.007595,-0.013702,-0.003935,-0.008290,-0.003921,0.007116,0.001854,0.006836,-0.005574,-0.007092,0.004808,-0.000243,0.004937,-0.004113
North-African,-0.072847,0.142174,0.003017,-0.072352,0.034776,-0.037371,-0.034312,0.008077,0.08999,0.032985,0.011367,-0.010191,0.027354,-0.016515,0.024701,-0.01485,0.001825,-0.026605,-0.053799,0.008004,-0.015847,-0.044515,0.020829,-0.012532,0.01449


[Душан]

Da, to je normalno, jer su rani Sloveni 65-70% Baltic + 30-35% Paleo-Balkan.

За нас на Балкану је релевантније да узимамо у обзир старословенску, а не балтичку компоненту, јер Словени приликом миграција на југ нису имали аутосомални профил Летонаца и Литванаца.

[Небојша]

За нас на Балкану је релевантније да узимамо у обзир старословенску, а не балтичку компоненту, јер Словени приликом миграција на југ нису имали аутосомални профил Летонаца и Литванаца.

Мислим да је Кочовић ишао логиком да су ти Early Slavic PCA позицијом релативно близу Paleo-Balkanic. Стога је у калкулатор ставио дијаметрално различите просеке, што је за G25 веома добро. Као што су Paleo-Balkanic и East-Med-Euro веома близу, па нема разлога да оба стоје у калкулатору.

Овде је данас најзаступљенији профил Paleo-Balkanic


Овде Baltic


Phoenician


Celtic

[kocovic]

За нас на Балкану је релевантније да узимамо у обзир старословенску, а не балтичку компоненту, јер Словени приликом миграција на југ нису имали аутосомални профил Летонаца и Литванаца.

Razumem tvoju logiku, istorijski si u pravu da Srbi iz 7-8. veka nisu identični modernim Letoncima. Međutim, ovde nije reč o istorijskoj simulaciji, već o matematičkoj stabilnosti G25 kalkulatora.

Problem sa Early-Slavic uzorcima je što su oni matematički 99.8% identični sa Baltic uzorcima. Ako u kalkulatoru imaš dve komponente koje su skoro iste, algoritam postaje nestabilan (možda si čuo za izraz na engleskom - overfitting). Biramo Baltic jer on predstavlja "čistiji" genetski pol te severne komponente.

Cilj je da imamo što manje preklapanje (overlap) između labela. Baltic nam služi kao "sidro" za severni drift, dok Paleo-Balkanic služi kao sidro za južni. Na taj način dobijamo najprecizniji uvid u to koliki je stvarni udeo slovenske, a koliki starobalkanske genetike, bez "šuma" koji bi unela Early-Slavic komponenta koja je i sama često blago pomerena ka jugu.

[kocovic]

Kapiram da objašnjenje gore možda nije jasno svakom pa sam odlučio da dam primer sa 2 koordinate (umesto 25 koliko koristi G25). Evo slike:


Dakle pošto je ugao između dve koordinate skoro pa nula to znači da su Early-Slavic i Baltic maltene iste populacije, samo je u pitanju odnos osnovnih populacija (iz kamenog doba).
To se dobija kosinusom ugla između 2 vektora.

Sada za prave G25 zamislite da svaki vektor ima 25 koordinate umesto 2, i taj ugao je skoro pa nula za Early-Slavic i Baltic.

Zato uzimamo samo onu čistiju (Baltic).

[Кековац]

Kapiram da objašnjenje gore možda nije jasno svakom pa sam odlučio da dam primer sa 2 koordinate (umesto 25 koliko koristi G25). Evo slike:


Dakle pošto je ugao između dve koordinate skoro pa nula to znači da su Early-Slavic i Baltic maltene iste populacije, samo je u pitanju odnos osnovnih populacija (iz kamenog doba).
To se dobija kosinusom ugla između 2 vektora.

Sada za prave G25 zamislite da svaki vektor ima 25 koordinate umesto 2, i taj ugao je skoro pa nula za Early-Slavic i Baltic.

Zato uzimamo samo onu čistiju (Baltic).

Да, али када се добије 25 координата, треба проверити угао између сваке две координате и видети да ли је тај угао мали као за ове две координате које си приказао на слици. То је 3.000 различитих парова. Очекивао бих да се нађе барем неки пар који показује разлику.

[ⲥуⲛцⲉ]

Target: G25sunce
Distance: 1.4647% / 0.01464740

51.4 Paleo-Balkanic
37.2 Baltic
5.4   Phoenician
5.2   Celtic
0.8   Uralic

[Небо_Сав]

Kapiram da objašnjenje gore možda nije jasno svakom pa sam odlučio da dam primer sa 2 koordinate (umesto 25 koliko koristi G25). Evo slike:


Dakle pošto je ugao između dve koordinate skoro pa nula to znači da su Early-Slavic i Baltic maltene iste populacije, samo je u pitanju odnos osnovnih populacija (iz kamenog doba).
To se dobija kosinusom ugla između 2 vektora.
Sada za prave G25 zamislite da svaki vektor ima 25 koordinate umesto 2, i taj ugao je skoro pa nula za Early-Slavic i Baltic.
Zato uzimamo samo onu čistiju (Baltic).

Rekao bih da G25 kalkulatori uvek nađu neko od mogućih rešenja, jer je algoritam grantovano konvergentan (nadje se zbir kvadrata razlika po koordinatama i izvadi koren iz tog zbira kvadrata) -on nalazi minimum euklidovske distance, i sam odredjuje to jedno iz mnoštva mogućih rešenja,  nego  je pitanje koliko je to rešenje smisleno i koliko  realno odražava komponente u uzorku.  Nije mi se do sada desilo da kalkulator prijavi problem singularnosti.

 Ukratko, ovaj G25 kalkulator radi kao i svaka gatara:  koju god kartu  da izvučeš - ona ima priču i za to, nema problema, to je samo pitanje njene imaginacije (i naše pshihologije)   

Druga je stvar sa F4 statistikama  i qpAdm analizom, koji su rigorozniji, i imaju kriterijum greške, pa ako niste prošli - idete u iteracije.

[kocovic]

Rekao bih da G25 kalkulatori uvek nađu neko od mogućih rešenja, jer je algoritam grantovano konvergentan (nadje se zbir kvadrata razlika po koordinatama i izvadi koren iz tog zbira kvadrata) -on nalazi minimum euklidovske distance, i sam odredjuje to jedno iz mnoštva mogućih rešenja,  nego  je pitanje koliko je to rešenje smisleno i koliko  realno odražava komponente u uzorku.  Nije mi se do sada desilo da kalkulator prijavi problem singularnosti.

 Ukratko, ovaj G25 kalkulator radi kao i svaka gatara:  koju god kartu  da izvučeš - ona ima priču i za to, nema problema, to je samo pitanje njene imaginacije (i naše pshihologije)   

Druga je stvar sa F4 statistikama  i qpAdm analizom, koji su rigorozniji, i imaju kriterijum greške, pa ako niste prošli - idete u iteracije.

G25 samo radi PCA preslikavanje vaših alela u koordinatu sa 25 komponenti. Taj koordinatni sistem ima 25 osa.

Pitanje je bilo zašto Baltic a ne Early-Slavic? Ja sam objasnio da oba vektora maltene pokazuju u istom smeru, što znači da su maltene kolinearni (mogu se konstruisati na istoj liniji, ali im se magnitude možda razlikuju), a pošto Baltic sadrži "antičkije" komponente njega biramo.

Da, naravno da je ADMIXTOOLS neuporedivo bolji jer je zasnovan na f-statistici (frekvenciji korelacije alela).

Nemoj da se zbuniš jer nakon PCA/f-statistike oba modela se baziraju na linearnoj algebri (mada različitim delovima - PCA se zasniva na geomterijskom modelovanju, a f-statistika na faktorizaciji matrice).

[kocovic]

Да, али када се добије 25 координата, треба проверити угао између сваке две координате и видети да ли је тај угао мали као за ове две координате које си приказао на слици. То је 3.000 различитих парова. Очекивао бих да се нађе барем неки пар који показује разлику.

Pomešao si neke stvari. Ima smisla porediti samo iste komponente, recimo PC 1 od Baltic i PC1 od Early-Slavic. Ako porediš PC3 Baltic i PC1 Early-Slavic efektivno porediš babe i žabe.

[Кековац]

Pomešao si neke stvari. Ima smisla porediti samo iste komponente, recimo PC 1 od Baltic i PC1 od Early-Slavic. Ako porediš PC3 Baltic i PC1 Early-Slavic efektivno porediš babe i žabe.

Наравно да нема, али ниси ни демонстирао поређење PC1 од Baltic и PC1 од Early-Slavic, него позиција Baltic и Early-Slavic кластера пројектоване на PC1 x PC2. Ја кажем да то што се та два кластера налазе на приближно истом месту на PC1 x PC2 равни не значи да ће се наћи на приближно истом месту на PC21 x PC22 равни.

[kocovic]

Наравно да нема, али ниси ни демонстирао поређење PC1 од Baltic и PC1 од Early-Slavic, него позиција Baltic и Early-Slavic кластера пројектоване на PC1 x PC2. Ја кажем да то што се та два кластера налазе на приближно истом месту на PC1 x PC2 равни не значи да ће се наћи на приближно истом месту на PC21 x PC22 равни.

Ne, nisam poredio PC1 i PC2. Poredio sam vektore pravca/trenda za Baltic i Early-Slavic. Kosinusna sličnost određuje trend, tj. da li su vektori kolinearni, upravni itd, kao i orijentaciju, a sve u jednom broju.

[Кековац]

Ne, nisam poredio PC1 i PC2. Poredio sam vektore pravca/trenda za Baltic i Early-Slavic. Kosinusna sličnost određuje trend, tj. da li su vektori kolinearni, upravni itd, kao i orijentaciju, a sve u jednom broju.

Ови вектори су колико видим са слика у PC1/PC2 равни. Моје питање је било шта нам гарантује да ће 25-димензионалном простору косонусна сличност бити приближно иста као у дводимензионалном?

[kocovic]

Ови вектори су колико видим са слика у PC1/PC2 равни. Моје питање је било шта нам гарантује да ће 25-димензионалном простору косонусна сличност бити приближно иста као у дводимензионалном?

Napisao sam da je to primer (jer kako da nacrtam 25-dimenzionalni vektor?). Ali 2 ili 3 dimenzije se mogu nacrtati. Poena je da ako ubacis u formulu iz trigonometrije: a•b = |a|•|b| cos(a, b) dobiješ kosinusnu sličnost dva vektora. Ovde su ti a i b vektori, a |a| i |b| magnitude vektora a i b. Ta formula ti daje ugao izmedju dva vektora, tj. da li pokazuju u istom smeru.

Dalje ove vektore (source) uzimamo kao koordinatne ose za računanje uzoraka od značaja (target). Ako imaš dve koordinatne ose koje su maltene u istoj liniji, razmisli da li ima smisla da ih koristiš?

[Небо_Сав]

Napisao sam da je to primer (jer kako da nacrtam 25-dimenzionalni vektor?).

Ne može kao 25-dimenzionalni vektor, ali može se videti  kao koordinatni profil, koji jeste vizualizacija 25D vektora kao linijskog grafikona.
 I ako se pažljivo gleda može se videti da su profili skoro identični na svih 25 dimanzija, a ne samo na prve dve. Ovo je preglednije nego slika u 2D  ili 3D prostoru.
Od toga smo i krenuli: 

Na primer:
PC                   1          2            3              4             5           6              7              8           ...
Germanic,0.123926,0.130753,0.063643,0.053182,0.038853,0.021314,0.008262,0.004974,0.002449,-0.003866,-0.004535,0.004916,-0.009130,-0.005401,0.016694,0.006442,-0.003732,0.00358
EarlySlavic,0.129471,0.121707,0.070142,0.064739,0.038239,0.024348,0.007002,0.012762,-0.001401,-0.021031,-0.001969,-0.006265,0.015024,0.020210,-0.009954,0.001262,0.005752,-0.000

[Небо_Сав]

Evo uporednog pregleda nekoliko uzoraka iz genetskog roda Ozrinića na osnovu novododatog kalkulatora:


https://www.exploreyourdna.com/calculator/216/early-european-calculator-by-sandr

Uzorci sa majkama iz istočnog dela današnje Crne Gore (Kuči i Plav) imaju veći udeo Paleo-Balkanic:

Moj grafikon je obeležen plavim pravouganikom.