Manual filtering of buffered genes

[zhiyi1988]

Hi there,

Thank you very much for programming DeltaTE. It is very helpful for studying translation regulation.

Recently, I ran an analysis and got some buffered genes, but it is confusing why they were in the "buffered" category and I'm wondering if the results should be further filtered?

Part of my results of buffered genes are as follows,

<style> </style>

	RNA_log2FC	p-value	Ribo_log2FC	p-value	TE_log2FC	p-value
gene1	-2.9	sig.	0.2	not sig.	3.4	sig.
gene2	1.2	sig.	0.2	not sig.	6	sig.
gene3	-0.1	sig.	-2.2	sig.	27.5	sig.

For the gene1, it is easy to understand that its transcription has been buffered, as higher TE resulted in unchanged translation level.

However, for gene2, although it is buffered according to the definition (i.e. significant delta-TE and delta-RNA, but non-significant delta-Ribo), the direction of delta-TE and delta-RNA is the same. It is hard to understand when a buffered gene's transcription and TE is higher under condition 1, but the translation level is not changed between two conditions. It seems that change in TE is not counteracting change in RNA. Maybe "delta-TE and delta-RNA are in opposite direction" is still the premise when defining buffered gene if the delta-Ribo is not significant?

Gene3 is also surely a buffered gene according to the definition, but is seems more like a intensified gene as delta-Ribo is larger than delta-RNA. In another words, if change in TE is counteracting change in RNA, won't delta-Ribo be smaller than delta-RNA?

Any advice would be much appreciated.

Best regards,
zhiyi

[soniachothani]

Thank you for your kind words and glad deltaTE has been helpful.

Theoretically, the calculation of the log fold change of TE should be ~ log fold change (Ribo2/1) - log fold change (RNA 2/1). You can see details in our paper in the section: Mathematical proof: Interaction term coefficient is equivalent to the changes in translation efficiency. So, I am quite surprised by the combination of Ribo-seq, RNA-seq and TE fold changes you observe in these two genes. To help you investigate this further, I would need more information. Could you share with me your PCA plots, the library depth for your Ribo-seq data and the Ribo-seq/RNA-seq log fold change scatter plot generated by deltaTE.

Generally, the guidelines mentioned in the paper in the section "STRATEGIC PLANNING" are to be followed and unreliable results can arise from less than three replicates or very low depth. "At least three biological replicates per condition or group are recommended for robust analysis of differential transcription, translation, and translational efficiency." and "Despite the presence of an experimental step to remove ribosomal RNA (rRNA) fragments from the input RNA, sequenced Ribo-seq reads still include a fraction of rRNA sequences, which should be discarded before ΔTE analysis. Thus, it is recommended to sequence at least 20 million reads per sample."

Best wishes,
Sonia