Tuberculosis (TB) is one of the most devastating chronic infectious diseases, but the role of host genetics in disease development after infection in this disease remains unidentified. Genome-wide association studies (GWASs) in Thais and Japanese were carried out and separately analyzed, attempted replication, then, combined by meta-analysis were not yielding any convincing association evidences; these results suggested that moderate to high effect-size genetic risks are not existed for TB per se. Because of failure in replication attempt of the top 50 single-nucleotide polymorphisms (SNPs) identified form meta-analysis data, we empirically split TB cases into young TB case/control data sets (GWAS-T(young)=137/295 and GWAS-J(young)=60/249) and old TB case/control data sets (GWAS-T(old)=300/295 and GWAS-J(old)=123/685), re-analyzed GWAS based on age-stratified data and replicated the significant findings in two independent replication samples (young TB; Rep-T(young)=155/249, Rep-J(young)=41/462 and old TB; Rep-T(old)=212/187, Rep-J(old)=71/619). GWAS and replication studies conducted in young TB identified at-risk locus in 20q12. Although the locus is located in inter-genic region, the nearest genes (HSPEP1-MAFB) from this locus are promising candidates for TB susceptibility. This locus was also associated with anti-TNF responsiveness, drug with increased susceptibility for TB. Moreover, eight SNPs in an old TB meta-analysis and six SNPs in young TB meta-analysis provided replication evidences but did not survive genome-wide significance.These findings suggest that host genetic risks for TB are affected by age at onset of TB, and this approach may accelerate the identification of the major host factors that affect TB in human populations.