The origin of tetrapod has been one of intense debating open questions for decades between coelacanth(Latimeria chalumnae) and lungfish (Protopterus annectens). For resolving this incongruence in phylogenies, a genome-wide data mining approach is used to retrieve 43 shared genes of seven taxa from GenBank and further 1001 orthologous genes of ten taxa from the Ensembl and NCBI. We used the maximum gene-support tree approach and the majority-rule branch approach to analyze 43 nuclear genes encoding amino acid residues and compared these results to those inferred with the concatenation approach. Our results successfully provide strong evidence in favor of the lungfish-tetrapod hypothesis, but rejecting the coelacanth-tetrapod hypothesis based on significantly fewer gene supports and lower taxon jackknife probabilities for the coelacanth-tetrapod clade than the lungfish-tetrapod one with the maximum gene-support tree approach and the jackknife method for taxon subsampling. When more and more genomic data become available in recent years, sequence data of 1001 shared genes was mined. We used the maximum gene-support approach with this larger dataset successfully to infer that lungfish is the closest relative of land vertebrates with a significant difference at p < 0.01 (Chi-Square test) in gene support values between a maximum gene-support tree and the second most gene support tree with ML methods. The second most support to the maximum (SM ratio), a relative value, is a better support index than a single absolute value of support to show the insight of the phylogenetic support. Our results also show increasing the number of shared genes is much more effective than increasing the number of taxa.