Assuming that there's no empty first field:

awk 'BEGIN{OFS=FS="\t";first=""}{if(NF==4){first=$1;print $0}else{print first,$0}}' input > output

If first field is empty:

awk 'BEGIN{OFS=FS="\t";first=""}{if($1!=""){first=$1; print $0}else{print first,$2,$3,$4}}' input > output

output using sed and awk:

$ sed 's/^pfam/\tpfam/g' test.txt |awk -F "\t" -v OFS="\t" '{if($1=="") {$1=previous} previous=$1}1'  
g1  pfam    PF12    ABC transporter
g2  pfam    PF13    transcription factor
g3  pfam    PF14    glycosyl hydrolase
g3  pfam    PF15    FAD binding domain  
g4  pfam    PF16    RTA1 like protein
g4  pfam    PF17    Zinc-binding dehydrogenase  
g4  pfam    PF18    major facilitator superfamily   
g5  pfam    PF19    short chain dehydrogenase
g6  pfam    PF20    ABC transporter

input (tab separated):

$ cat test.txt 
g1  pfam    PF12    ABC transporter
g2  pfam    PF13    transcription factor
g3  pfam    PF14    glycosyl hydrolase
pfam    PF15    FAD binding domain  
g4  pfam    PF16    RTA1 like protein
pfam    PF17    Zinc-binding dehydrogenase  
pfam    PF18    major facilitator superfamily   
g5  pfam    PF19    short chain dehydrogenase
g6  pfam    PF20    ABC transporter

This should work in python as long as you replace the "input.txt" and "output.txt" with the paths to your input file and desired output file.

f = open("input.txt", "r")
o = open("output.txt", "w")

lists = []

for line in f.readlines():
        temp_list = []
        index = 3
        if line.startswith("p"):
                temp_list.append(prev_g)
                index = 2
        for j in line.split()[:index]:
                temp_list.append(j)
        temp_list.append(" ".join(line.split()[index:]))
        lists.append(temp_list)
        if line.startswith("g"):
                prev_g = line.split()[0]

for i in lists:
        for j in i:
                o.write(str(j))
                o.write("\t")
        o.write("\n")