Most RNA-seq datasets involve the use of random hexamers primers for the amplification of RNAs present in samples. The use of such primers can lead to the presence of biases in the 5'-end region of the raw sequences.
Quality control tools such as Cutadapt, fastp and trimmomatic provide parameter options to remove such biased regions from raw sequences.