pyreml 0.0.0__py3-none-any.whl

pyreml/__init__.py

@@ -0,0 +1,3 @@
+from .larix import larix
+from .kinships import A_genomic, A_pedigree, D_pedigree
+from .nearPD import nearPD

pyreml/data/Larix.csv

@@ -0,0 +1,208 @@
+X,Y,ID,BLOC,VARIETY,DAM,SIRE,year,height,circumference,flexuosity
+16,32,16:32,B9,F0910,104.0,109.0,2000,475.0,150.0,4.0
+39,39,39:39,B11,F0910,104.0,109.0,2000,500.0,158.0,4.0
+35,68,35:68,B12,F0910,104.0,109.0,2000,472.0,160.0,5.0
+25,30,25:30,B15,F0910,104.0,109.0,2000,500.0,186.0,4.0
+36,55,36:55,B8,F0910,104.0,109.0,2000,549.0,192.0,4.0
+8,33,8:33,B5,F0910,104.0,109.0,2000,623.0,227.0,4.0
+31,32,31:32,B16,F0910,104.0,109.0,2000,597.0,238.0,4.0
+16,56,16:56,B4,F0910,104.0,109.0,2000,530.0,241.0,4.0
+66,50,66:50,B1,F0910,104.0,109.0,2000,617.0,265.0,4.0
+27,39,27:39,B7,F0910,104.0,109.0,2000,670.0,285.0,5.0
+34,70,34:70,B12,F0912,106.0,109.0,2000,348.0,105.0,5.0
+60,72,60:72,B13,F0912,106.0,109.0,2000,416.0,119.0,5.0
+8,36,8:36,B5,F0912,106.0,109.0,2000,424.0,135.0,4.0
+35,6,35:6,B10,F0912,106.0,109.0,2000,448.0,150.0,5.0
+34,27,34:27,B16,F0912,106.0,109.0,2000,500.0,162.0,4.0
+37,47,37:47,B11,F0912,106.0,109.0,2000,469.0,165.0,5.0
+48,31,48:31,B6,F0912,106.0,109.0,2000,456.0,167.0,4.0
+51,30,51:30,B2,F0912,106.0,109.0,2000,470.0,171.0,4.0
+27,45,27:45,B7,F0912,106.0,109.0,2000,515.0,178.0,4.0
+19,56,19:56,B4,F0912,106.0,109.0,2000,480.0,180.0,4.0
+54,74,54:74,B3,F0912,106.0,109.0,2000,497.0,181.0,3.0
+35,14,35:14,B14,F0912,106.0,109.0,2000,510.0,198.0,5.0
+36,57,36:57,B8,F0912,106.0,109.0,2000,603.0,215.0,4.0
+34,73,34:73,B12,F0914,106.0,166.0,2000,445.0,124.0,5.0
+62,50,62:50,B1,F0914,106.0,166.0,2000,459.0,143.0,3.0
+54,73,54:73,B3,F0914,106.0,166.0,2000,502.0,150.0,5.0
+38,52,38:52,B8,F0914,106.0,166.0,2000,500.0,150.0,5.0
+28,35,28:35,B15,F0914,106.0,166.0,2000,485.0,154.0,4.0
+59,72,59:72,B13,F0914,106.0,166.0,2000,508.0,158.0,4.0
+20,35,20:35,B9,F0914,106.0,166.0,2000,460.0,163.0,4.0
+36,7,36:7,B10,F0914,106.0,166.0,2000,459.0,166.0,3.0
+58,33,58:33,B2,F0914,106.0,166.0,2000,527.0,178.0,4.0
+51,36,51:36,B6,F0914,106.0,166.0,2000,540.0,190.0,3.0
+8,34,8:34,B5,F0914,106.0,166.0,2000,485.0,205.0,3.0
+39,48,39:48,B11,F0914,106.0,166.0,2000,555.0,220.0,4.0
+17,54,17:54,B4,F0914,106.0,166.0,2000,550.0,222.0,4.0
+58,69,58:69,B13,F0915,106.0,222.0,2000,320.0,120.0,5.0
+27,40,27:40,B7,F0915,106.0,222.0,2000,460.0,132.0,3.0
+35,15,35:15,B14,F0915,106.0,222.0,2000,393.0,144.0,5.0
+56,66,56:66,B3,F0915,106.0,222.0,2000,485.0,164.0,5.0
+40,51,40:51,B8,F0915,106.0,222.0,2000,500.0,165.0,5.0
+39,4,39:4,B10,F0915,106.0,222.0,2000,423.0,176.0,5.0
+59,33,59:33,B2,F0915,106.0,222.0,2000,476.0,190.0,4.0
+47,38,47:38,B6,F0915,106.0,222.0,2000,534.0,195.0,5.0
+16,36,16:36,B9,F0915,106.0,222.0,2000,522.0,204.0,5.0
+8,32,8:32,B5,F0915,106.0,222.0,2000,548.0,205.0,3.0
+19,55,19:55,B4,F0915,106.0,222.0,2000,567.0,222.0,4.0
+67,45,67:45,B1,F0915,106.0,222.0,2000,575.0,232.0,3.0
+37,43,37:43,B11,F0915,106.0,222.0,2000,512.0,232.0,4.0
+58,72,58:72,B13,F1002,106.0,221.0,2000,328.0,91.0,4.0
+50,75,50:75,B3,F1002,106.0,221.0,2000,380.0,108.0,5.0
+36,31,36:31,B16,F1002,106.0,221.0,2000,443.0,122.0,5.0
+39,38,39:38,B11,F1002,106.0,221.0,2000,355.0,130.0,3.0
+15,29,15:29,B9,F1002,106.0,221.0,2000,465.0,140.0,4.0
+41,54,41:54,B8,F1002,106.0,221.0,2000,430.0,166.0,4.0
+27,48,27:48,B7,F1002,106.0,221.0,2000,495.0,182.0,5.0
+64,42,64:42,B1,F1002,106.0,221.0,2000,498.0,215.0,5.0
+50,77,50:77,B3,F1015,242.0,222.0,2000,400.0,105.0,4.0
+37,4,37:4,B10,F1015,242.0,222.0,2000,452.0,152.0,3.0
+11,29,11:29,B5,F1015,242.0,222.0,2000,535.0,198.0,4.0
+64,43,64:43,B1,F1015,242.0,222.0,2000,605.0,218.0,3.0
+24,40,24:40,B7,F1042,104.0,214.0,2000,510.0,145.0,5.0
+22,29,22:29,B15,F1042,104.0,214.0,2000,555.0,170.0,5.0
+33,30,33:30,B16,F1042,104.0,214.0,2000,550.0,170.0,4.0
+59,31,59:31,B2,F1042,104.0,214.0,2000,547.0,250.0,3.0
+51,76,51:76,B3,F1118,104.0,106.0,2000,282.0,85.0,4.0
+40,75,40:75,B12,F1118,104.0,106.0,2000,390.0,136.0,5.0
+42,41,42:41,B11,F1118,104.0,106.0,2000,422.0,138.0,5.0
+11,33,11:33,B5,F1118,104.0,106.0,2000,452.0,148.0,4.0
+30,19,30:19,B14,F1118,104.0,106.0,2000,440.0,150.0,5.0
+24,47,24:47,B7,F1118,104.0,106.0,2000,470.0,158.0,4.0
+59,36,59:36,B2,F1118,104.0,106.0,2000,498.0,175.0,4.0
+68,74,68:74,B13,F1118,104.0,106.0,2000,445.0,178.0,5.0
+52,37,52:37,B6,F1118,104.0,106.0,2000,502.0,185.0,4.0
+38,11,38:11,B10,F1118,104.0,106.0,2000,509.0,188.0,5.0
+37,27,37:27,B16,F1118,104.0,106.0,2000,472.0,200.0,4.0
+58,50,58:50,B1,F1118,104.0,106.0,2000,543.0,218.0,4.0
+18,27,18:27,B9,F1118,104.0,106.0,2000,570.0,218.0,4.0
+13,50,13:50,B4,F1118,104.0,106.0,2000,570.0,221.0,4.0
+38,55,38:55,B8,F1118,104.0,106.0,2000,571.0,224.0,4.0
+23,28,23:28,B15,F1118,104.0,106.0,2000,555.0,242.0,4.0
+50,76,50:76,B3,F1119,104.0,284.0,2000,351.0,105.0,4.0
+41,71,41:71,B12,F1119,104.0,284.0,2000,466.0,152.0,5.0
+33,14,33:14,B14,F1119,104.0,284.0,2000,475.0,158.0,5.0
+56,30,56:30,B2,F1119,104.0,284.0,2000,475.0,160.0,4.0
+18,31,18:31,B9,F1119,104.0,284.0,2000,540.0,183.0,4.0
+67,77,67:77,B13,F1119,104.0,284.0,2000,460.0,186.0,4.0
+38,10,38:10,B10,F1119,104.0,284.0,2000,485.0,192.0,5.0
+58,40,58:40,B6,F1119,104.0,284.0,2000,580.0,195.0,3.0
+18,51,18:51,B4,F1119,104.0,284.0,2000,555.0,196.0,3.0
+23,30,23:30,B15,F1119,104.0,284.0,2000,525.0,216.0,4.0
+11,37,11:37,B5,F1119,104.0,284.0,2000,545.0,217.0,4.0
+63,72,63:72,B13,F1124,106.0,104.0,2000,338.0,105.0,5.0
+43,55,43:55,B8,F1124,106.0,104.0,2000,370.0,135.0,5.0
+50,30,50:30,B6,F1124,106.0,104.0,2000,474.0,174.0,4.0
+38,12,38:12,B10,F1124,106.0,104.0,2000,443.0,176.0,5.0
+56,64,56:64,B3,F1124,106.0,104.0,2000,494.0,183.0,4.0
+64,49,64:49,B1,F1124,106.0,104.0,2000,442.0,195.0,5.0
+6,27,6:27,B5,F1124,106.0,104.0,2000,565.0,199.0,5.0
+21,52,21:52,B4,F1124,106.0,104.0,2000,575.0,202.0,3.0
+37,30,37:30,B16,F1124,106.0,104.0,2000,520.0,216.0,4.0
+39,45,39:45,B11,F1124,106.0,104.0,2000,533.0,225.0,5.0
+18,29,18:29,B9,F1124,106.0,104.0,2000,560.0,238.0,4.0
+56,31,56:31,B2,F1125,106.0,284.0,2000,450.0,130.0,3.0
+30,20,30:20,B14,F1125,106.0,284.0,2000,423.0,153.0,4.0
+41,72,41:72,B12,F1125,106.0,284.0,2000,481.0,185.0,3.0
+23,27,23:27,B15,F1125,106.0,284.0,2000,515.0,197.0,4.0
+11,34,11:34,B5,F1125,106.0,284.0,2000,616.0,238.0,3.0
+32,30,32:30,B16,F1128,109.0,166.0,2000,560.0,207.0,4.0
+57,64,57:64,B3,F1129,109.0,221.0,2000,392.0,125.0,4.0
+38,16,38:16,B14,F1129,109.0,221.0,2000,362.0,170.0,4.0
+41,48,41:48,B11,F1129,109.0,221.0,2000,570.0,191.0,5.0
+52,73,52:73,B3,F1147,214.0,,2000,375.0,108.0,4.0
+14,29,14:29,B9,F1147,214.0,,2000,450.0,128.0,5.0
+35,51,35:51,B8,F1147,214.0,,2000,445.0,138.0,4.0
+42,38,42:38,B11,F1147,214.0,,2000,420.0,143.0,4.0
+23,39,23:39,B7,F1147,214.0,,2000,505.0,159.0,4.0
+22,50,22:50,B4,F1147,214.0,,2000,490.0,166.0,4.0
+59,49,59:49,B1,F1147,214.0,,2000,484.0,180.0,4.0
+60,27,60:27,B2,F1147,214.0,,2000,535.0,188.0,4.0
+31,13,31:13,B14,F1147,214.0,,2000,512.0,195.0,4.0
+26,27,26:27,B15,F1147,214.0,,2000,580.0,197.0,4.0
+34,28,34:28,B16,F1152,221.0,109.0,2000,545.0,218.0,4.0
+14,32,14:32,B9,F1153,221.0,109.0,2000,487.0,144.0,4.0
+10,34,10:34,B5,F1153,221.0,109.0,2000,545.0,205.0,4.0
+35,18,35:18,B14,F1153,221.0,109.0,2000,525.0,222.0,4.0
+40,73,40:73,B12,F1154,221.0,284.0,2000,484.0,140.0,4.0
+22,47,22:47,B7,F1154,221.0,284.0,2000,490.0,152.0,4.0
+63,69,63:69,B13,F1154,221.0,284.0,2000,493.0,166.0,5.0
+44,55,44:55,B8,F1154,221.0,284.0,2000,484.0,168.0,5.0
+57,65,57:65,B3,F1154,221.0,284.0,2000,505.0,172.0,4.0
+32,5,32:5,B10,F1154,221.0,284.0,2000,450.0,175.0,3.0
+37,40,37:40,B11,F1154,221.0,284.0,2000,545.0,190.0,5.0
+28,17,28:17,B14,F1154,221.0,284.0,2000,505.0,196.0,5.0
+14,35,14:35,B9,F1154,221.0,284.0,2000,495.0,198.0,4.0
+25,37,25:37,B15,F1154,221.0,284.0,2000,575.0,212.0,3.0
+54,39,54:39,B6,F1154,221.0,284.0,2000,487.0,214.0,4.0
+60,40,60:40,B6,F1160,242.0,106.0,2000,342.0,104.0,3.0
+38,14,38:14,B14,F1160,242.0,106.0,2000,385.0,130.0,4.0
+36,28,36:28,B16,F1160,242.0,106.0,2000,480.0,142.0,5.0
+38,8,38:8,B10,F1160,242.0,106.0,2000,443.0,158.0,4.0
+22,33,22:33,B15,F1160,242.0,106.0,2000,520.0,180.0,4.0
+61,50,61:50,B1,F1160,242.0,106.0,2000,556.0,185.0,3.0
+10,35,10:35,B5,F1160,242.0,106.0,2000,575.0,189.0,3.0
+68,75,68:75,B13,F1160,242.0,106.0,2000,500.0,195.0,4.0
+37,46,37:46,B11,F1160,242.0,106.0,2000,542.0,202.0,4.0
+23,45,23:45,B7,F1160,242.0,106.0,2000,615.0,207.0,3.0
+32,32,32:32,B16,F1162,242.0,221.0,2000,505.0,174.0,4.0
+42,43,42:43,B11,F1162,242.0,221.0,2000,550.0,235.0,5.0
+18,33,18:33,B9,F1164,242.0,,2000,411.0,101.0,3.0
+30,34,30:34,B16,F1164,242.0,,2000,435.0,132.0,5.0
+55,72,55:72,B3,F1164,242.0,,2000,465.0,137.0,4.0
+23,38,23:38,B7,F1164,242.0,,2000,500.0,152.0,3.0
+37,56,37:56,B8,F1164,242.0,,2000,473.0,154.0,5.0
+7,31,7:31,B5,F1164,242.0,,2000,527.0,161.0,4.0
+29,19,29:19,B14,F1164,242.0,,2000,438.0,172.0,4.0
+30,8,30:8,B10,F1164,242.0,,2000,492.0,180.0,5.0
+59,27,59:27,B2,F1164,242.0,,2000,578.0,202.0,4.0
+68,50,68:50,B1,F1164,242.0,,2000,588.0,214.0,4.0
+20,52,20:52,B4,F1164,242.0,,2000,590.0,220.0,4.0
+60,46,60:46,B1,F1267,104.0,214.0,2000,520.0,185.0,4.0
+46,50,46:50,B8,F1268,104.0,221.0,2000,355.0,125.0,5.0
+21,42,21:42,B7,F1268,104.0,221.0,2000,435.0,135.0,5.0
+36,68,36:68,B12,F1268,104.0,221.0,2000,460.0,152.0,5.0
+33,11,33:11,B10,F1268,104.0,221.0,2000,445.0,155.0,4.0
+68,72,68:72,B13,F1268,104.0,221.0,2000,403.0,171.0,4.0
+17,31,17:31,B9,F1268,104.0,221.0,2000,595.0,192.0,5.0
+39,43,39:43,B11,F1268,104.0,221.0,2000,568.0,208.0,5.0
+36,15,36:15,B14,F1268,104.0,221.0,2000,522.0,212.0,4.0
+10,29,10:29,B5,F1268,104.0,221.0,2000,602.0,243.0,4.0
+27,28,27:28,B15,F1268,104.0,221.0,2000,660.0,294.0,4.0
+20,32,20:32,B9,F1272,109.0,166.0,2000,460.0,130.0,4.0
+20,51,20:51,B4,F1272,109.0,166.0,2000,445.0,154.0,5.0
+37,39,37:39,B11,F1272,109.0,166.0,2000,626.0,206.0,4.0
+65,47,65:47,B1,F1272,109.0,166.0,2000,568.0,210.0,4.0
+53,28,53:28,B2,F1272,109.0,166.0,2000,560.0,214.0,4.0
+59,39,59:39,B6,F1275,109.0,222.0,2000,502.0,150.0,4.0
+58,74,58:74,B13,F1275,109.0,222.0,2000,455.0,155.0,5.0
+10,32,10:32,B5,F1275,109.0,222.0,2000,536.0,162.0,4.0
+41,73,KAMIE,B12,KAMIE,,,2000,284.0,90.0,3.0
+14,52,KAMIE,B4,KAMIE,,,2000,325.0,90.0,4.0
+21,35,KAMIE,B15,KAMIE,,,2000,405.0,110.0,4.0
+61,31,KAMIE,B2,KAMIE,,,2000,431.0,118.0,3.0
+32,28,KAMIE,B16,KAMIE,,,2000,362.0,120.0,3.0
+12,29,KAMIE,B9,KAMIE,,,2000,435.0,120.0,4.0
+52,67,KAMIE,B3,KAMIE,,,2000,478.0,131.0,4.0
+31,14,KAMIE,B14,KAMIE,,,2000,432.0,136.0,4.0
+54,34,KAMIE,B6,KAMIE,,,2000,410.0,138.0,3.0
+25,39,KAMIE,B7,KAMIE,,,2000,449.0,160.0,4.0
+11,35,KAMIE,B5,KAMIE,,,2000,488.0,164.0,3.0
+35,3,KAMIE,B10,KAMIE,,,2000,448.0,165.0,4.0
+34,51,KAMIE,B8,KAMIE,,,2000,528.0,226.0,4.0
+38,39,KAMIE,B11,KAMIE,,,2000,542.0,236.0,5.0
+34,34,RUDA ,B16,RUDA ,,,2000,335.0,80.0,5.0
+60,77,RUDA ,B13,RUDA ,,,2000,350.0,113.0,1.0
+38,70,RUDA ,B12,RUDA ,,,2000,388.0,121.0,5.0
+9,27,RUDA ,B5,RUDA ,,,2000,429.0,132.0,5.0
+48,28,RUDA ,B6,RUDA ,,,2000,425.0,132.0,4.0
+55,30,RUDA ,B2,RUDA ,,,2000,452.0,140.0,4.0
+35,10,RUDA ,B10,RUDA ,,,2000,410.0,150.0,3.0
+21,54,RUDA ,B4,RUDA ,,,2000,345.0,156.0,4.0
+50,67,RUDA ,B3,RUDA ,,,2000,492.0,162.0,5.0
+39,51,RUDA ,B8,RUDA ,,,2000,510.0,170.0,5.0
+26,46,RUDA ,B7,RUDA ,,,2000,570.0,180.0,3.0
+31,15,RUDA ,B14,RUDA ,,,2000,462.0,182.0,3.0
+59,50,RUDA ,B1,RUDA ,,,2000,502.0,210.0,3.0
+15,30,RUDA ,B9,RUDA ,,,2000,625.0,338.0,4.0

pyreml/kinships.py

@@ -0,0 +1,225 @@
+import numpy as np
+import pandas as pd
+
+def A_genomic(
+        X,
+        min_MAF: float = 0.05,
+        max_missing: float = 0.1,
+        shrink: bool = False
+    ) :
+    '''
+    This is a direct retranscription of the A_mat
+    function from the R paclage: rrBLUP
+    X is a diploid species SNP matrix encoded: -1, 0, 1
+    the following parameters are fixed:
+        - impute.method = "mean"
+        - shrink.method = "EJ"
+    '''
+
+    n = X.shape[0]
+
+    frac_missing = np.isnan(X).mean(axis=0)
+    freq = np.nanmean(X + 1, axis=0) / 2
+    MAF = np.minimum(freq, 1 - freq)
+
+    markers = np.where((MAF >= min_MAF) & (frac_missing <= max_missing))[0]
+    m = len(markers)
+    var_A = 2 * np.mean(freq[markers] * (1 - freq[markers]))
+    mono = np.where(freq * (1 - freq) == 0)[0]
+    X[:, mono] = 2 * freq[mono] - 1
+    freq_mat = np.ones((n, 1)) @ freq[markers].reshape(1, -1)
+    W = X[:, markers] + 1 - 2 * freq_mat
+
+    # imputation
+    W[np.isnan(W)] = 0
+
+    if shrink:
+        Z = W - W.mean(axis=1, keepdims=True)
+        Z2 = Z ** 2
+        S = (Z @ Z.T) / m
+        target = np.mean(np.diag(S)) * np.eye(n)
+        var_S = (Z2 @ Z2.T) / m**2 - S**2 / m
+        b2 = var_S.sum()
+        d2 = ((S - target) ** 2).sum()
+        delta = float(np.clip(b2 / d2, 0, 1))
+        print(f"Shrinkage intensity: {delta:.2f}")
+        cov_W = target * delta + (1 - delta) * S
+        W_mean = W.mean(axis=1)
+        K = (cov_W + np.outer(W_mean, W_mean)) / var_A
+    else:
+        K = (W @ W.T) / var_A / m
+
+    return K
+
+def A_pedigree(pedigree: pd.DataFrame) -> np.ndarray:
+    '''
+    This is a direct retranscription of the makeD
+    function from the R paclage: nadiv
+    pedigree is a pandas dataframe with 3 columns:
+        - ID: the individual
+        - DAM and SIRE: its parents, may be NA
+    '''
+    ped = pedigree.copy()
+    ped.columns = ped.columns.str.lower()
+
+    # --- checks ---
+    if not {"id", "dam", "sire"}.issubset(ped.columns):
+        raise ValueError("Pedigree must have columns: id, dam, sire")
+
+    if ped["id"].duplicated().any():
+        dups = ped["id"][ped["id"].duplicated()].unique().tolist()
+        raise ValueError(f"Duplicate individuals: {dups}")
+
+    known_dam = ped["dam"].dropna().unique()
+    known_sire = ped["sire"].dropna().unique()
+    all_ids = set(ped["id"])
+    missing_parents = (set(known_dam) | set(known_sire)) - all_ids
+    if missing_parents:
+        raise ValueError(f"Parents not in pedigree: {sorted(missing_parents)}")
+
+    # --- generation assignment + sort ---
+    ids = ped["id"].values
+    dam = ped["dam"].values
+    sire = ped["sire"].values
+    n = len(ids)
+    id_to_idx = {v: i for i, v in enumerate(ids)}
+
+    gen = np.zeros(n, dtype=int)
+    changed = True
+    while changed:
+        changed = False
+        for i in range(n):
+            g = 0
+            if pd.notna(dam[i]):
+                g = max(g, gen[id_to_idx[dam[i]]] + 1)
+            if pd.notna(sire[i]):
+                g = max(g, gen[id_to_idx[sire[i]]] + 1)
+            if g > gen[i]:
+                gen[i] = g
+                changed = True
+
+    # sort by (gen, dam, sire) with NA first (like na.last=FALSE in R)
+    dam_key = np.array([id_to_idx[d] + 1 if pd.notna(d) else 0 for d in dam])
+    sire_key = np.array([id_to_idx[s] + 1 if pd.notna(s) else 0 for s in sire])
+    sort_idx = np.lexsort((sire_key, dam_key, gen))
+
+    ids_s = ids[sort_idx]
+    dam_s = dam[sort_idx]
+    sire_s = sire[sort_idx]
+
+    # numeric pedigree (0-based indices in sorted order)
+    id_to_sidx = {v: i for i, v in enumerate(ids_s)}
+    dam_idx = np.array([id_to_sidx[d] if pd.notna(d) else n for d in dam_s])
+    sire_idx = np.array([id_to_sidx[s] if pd.notna(s) else n for s in sire_s])
+
+    # --- Meuwissen & Luo: f and dii ---
+    f = np.zeros(n + 1)
+    f[n] = -1.0
+    dii = np.zeros(n)
+    AN = np.empty(2 * n, dtype=float)
+    li = np.zeros(n)
+
+    for k in range(n):
+        dii[k] = 0.5 - 0.25 * (f[dam_idx[k]] + f[sire_idx[k]])
+
+        if k > 0 and dam_idx[k] == dam_idx[k-1] and sire_idx[k] == sire_idx[k-1]:
+            f[k] = f[k-1]
+        else:
+            li[k] = 1.0
+            ai = 0.0
+            j = k
+            cnt = 0
+            while j >= 0:
+                sj = sire_idx[j]
+                dj = dam_idx[j]
+                if sj < n:
+                    AN[cnt] = sj
+                    li[sj] += 0.5 * li[j]
+                    cnt += 1
+                if dj < n:
+                    AN[cnt] = dj
+                    li[dj] += 0.5 * li[j]
+                    cnt += 1
+                ai += li[j] * li[j] * dii[j]
+                j = -n - 1  # empty sentinel
+                for h in range(cnt):
+                    if AN[h] > j:
+                        j = int(AN[h])
+                for h in range(cnt):
+                    if AN[h] == j:
+                        AN[h] = -n - 1
+            f[k] = ai - 1.0
+            li[:k+1] = 0.0
+
+    # --- Tinv and assembly ---
+    Tinv = np.eye(n)
+    for i in range(n):
+        if dam_idx[i] < n:
+            Tinv[i, dam_idx[i]] = -0.5
+        if sire_idx[i] < n:
+            Tinv[i, sire_idx[i]] = -0.5
+
+    sqrt_dii_inv = np.diag(np.sqrt(1.0 / dii))
+    L = np.linalg.solve(sqrt_dii_inv @ Tinv, np.eye(n))
+    A = L @ L.T
+
+    # --- reorder to original pedigree order ---
+    inv_sort = np.argsort(sort_idx)
+    A = A[np.ix_(inv_sort, inv_sort)]
+
+    return A
+
+def D_pedigree(pedigree: pd.DataFrame) -> np.ndarray:
+    '''
+    This is a direct retranscription of the makeD
+    function from the R paclage: nadiv
+    pedigree is a pandas dataframe with 3 columns:
+        - ID: the individual
+        - DAM and SIRE: its parents, may be NA
+    
+    NB1: makeD is an approximation, it is only exact
+    in the absence of inbreeding.
+
+    NB2: whenever one or the other parent is missing,
+    the individual is treated as a founder (both
+    parents missing)
+    '''
+    A = A_pedigree(pedigree)
+
+    ped = pedigree.copy()
+    ped.columns = ped.columns.str.lower()
+
+    ids = ped["id"].values
+    dam = ped["dam"].values.copy()
+    sire = ped["sire"].values.copy()
+
+    # force founders when only one parent is known
+    only_dam = pd.notna(dam) & pd.isna(sire)
+    only_sire = pd.isna(dam) & pd.notna(sire)
+    dam[only_dam] = None
+    sire[only_sire] = None
+
+    id_to_idx = {v: i for i, v in enumerate(ids)}
+    n = len(ids)
+
+    dam_idx = np.array([id_to_idx[d] if pd.notna(d) else -1 for d in dam])
+    sire_idx = np.array([id_to_idx[s] if pd.notna(s) else -1 for s in sire])
+
+    D = np.zeros((n, n))
+    half_A = A / 2.0
+
+    for k in range(n):
+        if dam_idx[k] == -1:
+            continue
+        for j in range(k):
+            if dam_idx[j] == -1:
+                continue
+            dk, sk = dam_idx[k], sire_idx[k]
+            dj, sj = dam_idx[j], sire_idx[j]
+            val = half_A[dk, dj] * half_A[sk, sj] + half_A[dk, sj] * half_A[sk, dj]
+            if val != 0.0:
+                D[k, j] = val
+                D[j, k] = val
+
+    np.fill_diagonal(D, 2.0 - np.diag(A))
+    return D

pyreml/larix.py

@@ -0,0 +1,7 @@
+import pandas as pd
+import os
+
+_pkg_dir = os.path.dirname(__file__)
+_data_dir = os.path.join(_pkg_dir, "data")
+
+larix = pd.read_csv(os.path.join(_data_dir, "Larix.csv"))

pyreml/nearPD.py

@@ -0,0 +1,19 @@
+import torch
+import numpy as np
+
+def nearPD(S, eps: float = 1e-4):
+    """
+    Nearest positive-definite surrogate of a symmetric matrix.
+
+    Symmetrize, clamp the eigenvalues at `eps`, rebuild, symmetrize again.
+    Guarantees strictly positive eigenvalues, which is what the factor-analytic
+    initialization relies on (the q dominant eigenvalues feed Lambda > 0).
+    """
+    was_numpy = isinstance(S, np.ndarray)
+    S = torch.as_tensor(S, dtype=torch.double)
+    S = (S + S.T) / 2
+    eigvals, eigvecs = torch.linalg.eigh(S)
+    eigvals = torch.clamp(eigvals, min=eps)
+    S = (eigvecs * eigvals) @ eigvecs.T
+    S = (S + S.T) / 2
+    return S.numpy() if was_numpy else S

pyreml-0.0.0.dist-info/METADATA

@@ -0,0 +1,37 @@
+Metadata-Version: 2.4
+Name: pyreml
+Version: 0.0.0
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: torch
+Requires-Dist: numpy
+Requires-Dist: scipy
+Requires-Dist: pandas
+Requires-Dist: patsy
+Requires-Dist: pyyaml
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+
+# Pyreml
+
+**pyreml** is a general purpose mixed model solver under active development.
+
+## Installation
+
+```bash
+pip install pyreml
+```
+
+## Documentation
+
+Read the manual for further documentation.
+
+## License
+
+Copyright 2026 Alexandre Marchal, Daniele Raimondi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

pyreml-0.0.0.dist-info/RECORD

@@ -0,0 +1,9 @@
+pyreml/__init__.py,sha256=onOlg5kSyKO9nUUgN6RN8Og0Nu2dkyRAttNfwPfbXxQ,108
+pyreml/kinships.py,sha256=8kYUTwHjpfSvWskDsbCRHkVL_IuExHCjFSApvdFCNuE,6995
+pyreml/larix.py,sha256=v6FLFHCA2VuumesZ498Zdv5anHKHRjuaodba0xtsgzQ,169
+pyreml/nearPD.py,sha256=nhTi1k7960zEPDVVsXrKalr4NfmQn6h9PplH0i-Pr8w,678
+pyreml/data/Larix.csv,sha256=sB5TL2F0Lj4_orNZOW6mKsypHUP5Jprwp2FriO5OP5c,10914
+pyreml-0.0.0.dist-info/METADATA,sha256=cOVIxsHIVIK8qzU7M1GgT9B5MJPd4m0dMY0E1v93XWg,1597
+pyreml-0.0.0.dist-info/WHEEL,sha256=aeYiig01lYGDzBgS8HxWXOg3uV61G9ijOsup-k9o1sk,91
+pyreml-0.0.0.dist-info/top_level.txt,sha256=JTwN14yYKYIc8XQPqTyuBEPEXJWwM1wiqQ5PyyxIYv8,7
+pyreml-0.0.0.dist-info/RECORD,,

pyreml-0.0.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (82.0.1)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

pyreml-0.0.0.dist-info/top_level.txt

@@ -0,0 +1 @@
+pyreml