Algorithm 1 Boruta Algorithm

Input:$\mathcal{M}$: Genotype (rSNPs) data Input:$\mathcal{L}$: Labels (cultivars) Output:$\mathcal{C}$: A ranked list of rSNPs based on their importance score Method: 1: $t=0$ 2: repeat 3:     ${\mathcal{M}}_t=\mathcal{M}$ 4:     ${\widehat{\mathcal{M}}}_t=\mathtt{shuffle}\left({\mathcal{M}}_t\right)$: Creation of shadow attributes 5:     ${\mathcal{M}}_t^{ext}=[{\mathcal{M}}_t;{\widehat{\mathcal{M}}}_t;\mathcal{L}]$: Matrix (data) concatenation to extend the input data 6:     ${\mathcal{VIS}}_t\left({\mathcal{M}}_t^{ext}\right)=RF\left({\mathcal{M}}_t^{ext}\right)$: Gathering variable importance scores $\left(\mathcal{VIS}\right)$ using RF classifier 7:     ${\widehat{\mathcal{VIS}}}_t=\mathtt{max}\left(\mathcal{VIS}\left({\widehat{\mathcal{M}}}_t\right)\right)$: Max. $\mathcal{VIS}$ value (in terms of z-Score) for shadow attributes 8:     ${\mathcal{M}}_t^c={\mathcal{M}}_t^{ext}[\mathcal{VIS}\left({\mathcal{M}}_t^{ext}\right)>{\widehat{\mathcal{VIS}}}_t]\backslash {\widehat{\mathcal{M}}}_t$: rSNPs with significantly higher $\mathcal{VIS}$ values $>\widehat{\mathcal{VIS}}$ 9:     ${\mathcal{M}}_t^r={\mathcal{M}}_t^{ext}[\mathcal{VIS}\left({\mathcal{M}}_t^{ext}\right)<{\widehat{\mathcal{VIS}}}_t]\backslash {\widehat{\mathcal{M}}}_t$: rSNPs with significantly lower $\mathcal{VIS}$ values $<\widehat{\mathcal{VIS}}$ 10:     $\mathcal{M}={\mathcal{M}}_t\backslash [{\mathcal{M}}_t^c;{\mathcal{M}}_t^r]$: Remove all rSNPs with determined importance from the input ${\mathcal{M}}_t$ 11:     ${\mathcal{C}}_t=\mathcal{VIS}\left({\mathcal{M}}_t^c\right)$: Gathering the rSNPs with confirmed $\mathcal{VIS}$ 12:     $t=t+1$ 13: until importance of all rSNPs is assigned 14: $\mathcal{C}={\bigcup }_{i=1}^t{\mathcal{C}}_i$