Table 1.

Comprehensive list of the steps implemented by SumStatsRehab

Step #	Description	Stage
1	Check if the input file is a zip or a gzip archive. If so, extract the raw file	Formatting
2	Pick and reorder the columns to the internal format, based on the corresponding *.json config file
3	Cut the ‘chr’ prefix in the chromosome column, if present
4	Calculated the weighted average of EAF columns into one EAF column, if multiple were specified
5	Check if a chromosome entry is an integer from 1 to 23, or X, Y, M. If an entry is not, mark it as invalid	Validation
6	Check if a base pair position entry is a non-negative integer. If an entry is not, mark it as invalid
7	Check if an rsID entry is a non-negative integer with ‘r’ prefix. If an entry is not, mark it as invalid
8	Check if an effect allele or other allele entry is either a dash or composed of letters ATCG. If an entry is not, mark as invalid
9	Check if a p-value or an EAF entry is a real value between 0 and 1 inclusively. If an entry is not, mark as invalid
10	Check if a standard error or a beta entry is a real value. If an entry is not, mark as invalid
11	Calculate statistics and save the report about correctness of the data
12	Analyze the report, If no resolvable issues were found, finish the execution	Analysis and preparation
13	If resolvable issues were found, prepare the restoration algorithm depending on the issues
14	Perform the liftover to build 38 if needed
15	Sort the sumstats file either by ChrBP or by rsID, depending on the restoration algorithm
16	If a chromosome or a BP entry was marked as invalid, and the sumstats file is sorted by rsID, then restore both entries by a lookup in the dbSNP for matching rsID	Restoration
17	If an rsID entry was marked as invalid, and the sumstats file is sorted by ChrBP, then restore rsID entry by a lookup in the dbSNP for matching Chr and BP
18	If from EA and OA entries only one is invalid then restore the invalid allele as the most likely allele from known by a lookup in the dbSNP either for matching rsID and valid allele, if sumstats is sorted by rsID, or for matching Chr, BP, and valid allele, if sumstats is sorted by ChrBP
19	If an EAF entry was marked as invalid, then EAF is restored by a lookup in the dbSNP either for matching rsID and effect allele, if sumstats is sorted by rsID, or for a matching Chr, BP, and effect allele, if sumstats is sorted by ChrBP
20	If a standard error, beta, or p-value entry was marked as invalid, and the other two entries as valid, then restore the invalid using the formula s = β/z, where s is the standard error, β is beta, and z is z-score that corresponds to the p-value in the two-tailed test [10]
21	Go back to step #5 (Validation stage)