Riesensackteil
Member
- Joined
- Apr 4, 2021
- Messages
- 499
Discussion and outlook
"We examined published sequence data (BioProject accession number PRJNA603194 in the NCBI Sequence Read Archive (SRA) database) on the genome sequence for SARS-CoV-2 (GenBank: MN908947.3) using a simple bioinformatics approach. The methods we used are not specific to SARS-CoV-2 and can be applied to other sequence data without special modifications.
First, we repeated the contig generation with Megahit (v.1.2.9) using the available sequence data and obtained significantly different results compared to the representations in [1]. In particular, we were unable to reproduce the longest contig with a length of 30,474 nt, which according to [1] comprised almost the entire viral genome and acted as the basis for primer design. On the contrary, the longest contig we generated (29,802 nt) showed a nearly complete match with reference MN908947.3. Consequently, the published sequence data cannot be the original short reads used for contig generation. This is to be regarded as extremely problematic in the context of scientific publications, since in this way it is no longer possible to verify the published results. The possibility to verify published scientific hypotheses is the essence of living science.
Contrary to what was reported in [1], we may have found contigs with high coverage associated with (ribosomal) ribonucleic acids of human origin. Thus, it is possible that not all human-associated nucleic acids were eliminated in the construction of SARSCoV-2. Further, no evidence of the presence of viral nucleic acids in the patient sample was provided and, consequently, there is a possibility that human or nonviral nucleic acid fragments were used to construct the claimed viral sequence MN908947.3 to a significant extent without detection. This possibility would have to be excluded by control experiments.
In all publications on the reference genomes analyzed in this study, the necessary evidence on the exact origin of the sequence fragments used for construction was also not provided and the necessary control experiments were not published.
We would like to mention here that control experiments may have already been performed many times without being noticed, showing the possibility of constructing 19 SARS-CoV-2 genomes from non-infectious human samples. For example, whole genome sequencing from samples with a baseline Ct value greater than 35 is reported in [5] and [17]. This could be a refutation for the viral model for SARS-CoV-2.
The analysis of the nucleotide coverage distributions as well as the length distributions of the mappable sequence reads for the respective reference sequences leads to the hypothesis of a possible unintentional amplification of sequence reads not associated with SARS-CoV-2. Further, along with this, the possibility of accidental generation of sequences that were not present in the initial sample but were generated only by the amplification conditions, such as the primer sequences used and the cycles performed, must be considered. This possibility therefore requires the performance of appropriate control experiments.
In addition to attempting to replicate the assembly published in [1] with the published sequence reads, we considered a simple approach for analyzing the internal structure of large datasets of short sequence reads. With the sequence data at hand, we were able to compute consensus sequences for the reference genomes LC312715.1 (HIV) and NC_001653.2 (Hepatitis delta virus) with higher goodness than for those reference sequences we considered associated with coronaviruses. This was particularly true for bat-SL-CoVZC45 (GenBank: MG772933.1), which led to the origin hypothesis of SARS-CoV-2. Thus, we were able to substantiate our hypothesis that the claimed viral genome sequences are misinterpretations in the sense that they have been or are being constructed unnoticed from non-viral nucleic acid fragments. In particular, our results underscore the urgent need to perform appropriate control experiments. For each suspected pathogenic viral genome sequence, an obvious protocol would be to attempt assembly of the genome sequences from corresponding non-suspect samples using identical protocols.
We observed high R1 and R2 error rates in the reference genomes for measles, Ebola, or Marburg, where the nucleic acid fragments used for construction were propagated in Vero cells. It remains an open question so far whether this is due to the nucleic acid sources themselves, or to the amplification conditions used (e.g. primer sequences and cycle number) or sequencing protocols (e.g. the polymerases and reverse transcriptases used).
With regard to our results, in addition to publishing the final sequence data used, we always recommend publishing sequence data that resulted only from amplification with random hexamers and moderate cycle numbers to provide the most unbiased data possible for structural analysis."
"We examined published sequence data (BioProject accession number PRJNA603194 in the NCBI Sequence Read Archive (SRA) database) on the genome sequence for SARS-CoV-2 (GenBank: MN908947.3) using a simple bioinformatics approach. The methods we used are not specific to SARS-CoV-2 and can be applied to other sequence data without special modifications.
First, we repeated the contig generation with Megahit (v.1.2.9) using the available sequence data and obtained significantly different results compared to the representations in [1]. In particular, we were unable to reproduce the longest contig with a length of 30,474 nt, which according to [1] comprised almost the entire viral genome and acted as the basis for primer design. On the contrary, the longest contig we generated (29,802 nt) showed a nearly complete match with reference MN908947.3. Consequently, the published sequence data cannot be the original short reads used for contig generation. This is to be regarded as extremely problematic in the context of scientific publications, since in this way it is no longer possible to verify the published results. The possibility to verify published scientific hypotheses is the essence of living science.
Contrary to what was reported in [1], we may have found contigs with high coverage associated with (ribosomal) ribonucleic acids of human origin. Thus, it is possible that not all human-associated nucleic acids were eliminated in the construction of SARSCoV-2. Further, no evidence of the presence of viral nucleic acids in the patient sample was provided and, consequently, there is a possibility that human or nonviral nucleic acid fragments were used to construct the claimed viral sequence MN908947.3 to a significant extent without detection. This possibility would have to be excluded by control experiments.
In all publications on the reference genomes analyzed in this study, the necessary evidence on the exact origin of the sequence fragments used for construction was also not provided and the necessary control experiments were not published.
We would like to mention here that control experiments may have already been performed many times without being noticed, showing the possibility of constructing 19 SARS-CoV-2 genomes from non-infectious human samples. For example, whole genome sequencing from samples with a baseline Ct value greater than 35 is reported in [5] and [17]. This could be a refutation for the viral model for SARS-CoV-2.
The analysis of the nucleotide coverage distributions as well as the length distributions of the mappable sequence reads for the respective reference sequences leads to the hypothesis of a possible unintentional amplification of sequence reads not associated with SARS-CoV-2. Further, along with this, the possibility of accidental generation of sequences that were not present in the initial sample but were generated only by the amplification conditions, such as the primer sequences used and the cycles performed, must be considered. This possibility therefore requires the performance of appropriate control experiments.
In addition to attempting to replicate the assembly published in [1] with the published sequence reads, we considered a simple approach for analyzing the internal structure of large datasets of short sequence reads. With the sequence data at hand, we were able to compute consensus sequences for the reference genomes LC312715.1 (HIV) and NC_001653.2 (Hepatitis delta virus) with higher goodness than for those reference sequences we considered associated with coronaviruses. This was particularly true for bat-SL-CoVZC45 (GenBank: MG772933.1), which led to the origin hypothesis of SARS-CoV-2. Thus, we were able to substantiate our hypothesis that the claimed viral genome sequences are misinterpretations in the sense that they have been or are being constructed unnoticed from non-viral nucleic acid fragments. In particular, our results underscore the urgent need to perform appropriate control experiments. For each suspected pathogenic viral genome sequence, an obvious protocol would be to attempt assembly of the genome sequences from corresponding non-suspect samples using identical protocols.
We observed high R1 and R2 error rates in the reference genomes for measles, Ebola, or Marburg, where the nucleic acid fragments used for construction were propagated in Vero cells. It remains an open question so far whether this is due to the nucleic acid sources themselves, or to the amplification conditions used (e.g. primer sequences and cycle number) or sequencing protocols (e.g. the polymerases and reverse transcriptases used).
With regard to our results, in addition to publishing the final sequence data used, we always recommend publishing sequence data that resulted only from amplification with random hexamers and moderate cycle numbers to provide the most unbiased data possible for structural analysis."
