How to improve the acurracy in training?

[bioinformaticspcj]

Dear authors, I am trying to train a model for species Thamnophis elegans, but found it very difficult to improve the gene level sensitivity which was always <0.2. I have trained the model with 4000 genes in a gff file from NCBI. I filtered the genes withount "start codon"and " stop codon", and genes with "in frame codon" were also filtered. All genes used are less similar than 0.7 among each others. I have no idea how to improve the gene level sensitivity.The command is as follows:

"perl augustus-3.3.1/scripts/autoAugTrain.pl --optrounds=5 --trainingset=GCF_009769535.1_rThaEle1.pri_genomic.sub.good.gff --genome=GCF_009769535.1_rThaEle1.pri_genomic.fna --species=Terrestrial_Garter --useexisting --workingdir=Tele"

I am looking forward to your timely reply and helpful suggestion. Thanks a lot. Best,

[linshengnan09]

hello, do you sovle this problem?