Conversion to exponential format of chromosomal location causing " --extract range" problems

[Daniel Felsky]

Hello!

I have imputed data using default parameters on impute2 (5mb chunks across autosome = ~556 total successful imputed chunks) and then used the fcgene program (http://sourceforge.net/projects/fcgene/) to convert each chunk into .bim/.bed/.fam format, excluding imputed variants with info score below 0.5. I then performed a simple operation using PLINK (v1.90b, updated today, Dec 29 2014) on each chunk as follows (input names changed for simplicity):

plink --bfile input --extract list.txt --make-bed --out output_extracted

Strangely, the chromosomal coordinates column in the output .bim files (in this case would be output_extracted.bim) have been converted into exponent form (i.e. 104046834 is now 1.0404683e+08). After a subsequent large-scale merge that combines all 556 sets of binary PLINK-format files, the same column in the now whole-genome output is shortened further (i.e. 1.0404683e+08 is now 1.04047e+08). As a result, when I try to extract SNP data form the whole-genome file over a specific range (using --extract range: plink --bfile IMPUTED_DATA --extract range chrom_interval.txt --recodeAD --out interval_extracted), PLINK throws an error: "Error: All variants excluded by '--extract range'." , despite the fact that my bp range of 103996234-104184834 (my "chrom_interval.txt" file has one line: "14 103996234 104184834 R1") includes my SNP of interest that I know exists in the dataset.

Thank you!

-Dan

[Christopher Chang]

What do the first two lines of the fcgene-generated .bim file look like?  Normally, a centimorgan coordinate (or, more commonly, just '0') is in the 3rd column, a bp coordinate should be in the fourth column, and minor/major alleles should be in the 5th and 6th.