MIT recently concluded a multi-year investigation into allegations of research misconduct lodged at Professor Ram Sasisekharan and other members of the Sasisekharan Lab, finding that none of the allegations were supported. Allegations of possible research misconduct in the Sasisekharan Lab appeared in the media following the abrupt publication of a perspective piece. MIT simultaneously received an internal complaint raising formal allegations of research misconduct against Sasisekharan Lab concerning the same research discussed in the perspective. Based on a comprehensive review of the evidentiary record, MIT concluded its investigation with no finding of research misconduct for any of the submitted allegations. MIT will assist in fully restoring the reputations of Professor Sasisekharan and the members of his lab. In addition, MIT is committed to ensuring that Professor Sasisekharan and Sasisekharan Lab can return to normal operation.
MIT’s Statement can be found here.
Professor Sasisekharan’s Statement on behalf of the Sasisekharan Lab can be found here.
The Science Adimab and Others Willfully Ignored
The Attack:
In May 2019, the Perspective “Connecting the Sequence Dots” was published in the online journal mAbs by authors from Adimab, LLC, including its co-founder (and CEO at the time of the publication) Tillman Gerngross and co-founder, MIT Professor Dane Wittrup (a member of the mAbs Editorial board) and Laura Walker fromer CSO of Adagio a/k/a INVIVYD, INC., (a Adimab Spinoff). The mAbs paper was written to imply that two antibodies VIS410 and ZAb_FLEP, which were respectively reported in peer-reviewed publications in PNAS in 2015 and Cell Host Microbe in 2018 (CHM paper) by the Ram Sasisekharan Lab at MIT, were plagiarized versions of the previously known antibodies FI6 (Humabs later acquired by Vir Biotechnology) and EDE1 C8 respectively. The mAbs paper further implied that VIS410 and ZAb_FLEP —each of were designed by introducing inconsequential mutations respectively on FI6 and EDE1 C8 for the purpose of hiding their origins and that MIT’s Sasisekharan Lab falsified their methods to hide the alleged plagiarism.
The publication process for the mAbs paper was quite unusual because:
- It was accepted within two days of receipt of the manuscript by mAbs where Dane Wittrup served on the editorial board.
- A pre-print of this article was given to the Wall Street Journal whose reporter then provided it to Professor Sasisekharan on May 6, 2019, which is when and how Professor Sasisekharan first learned of the “scientific” questions his “colleagues” had regarding Sasisekharan Lab antibody development methods.
- Neither Dane Wittrup nor any other faculty within MIT reached out to Ram Sasisekharan with questions or issues regarding the science that was attacked in the article.
Therefore, there were factors beyond healthy discussion and debate of science at play. The very public attack Sasisekharan Lab that endured, together with Sasisekharan Lab’s inability to defend these allegations publicly because of MIT Policy 10.1.5 governing research misconduct, was a most egregious abuse of the research misconduct process. Simultaneous publication of the mAbs Perspective together with the submission of research misconduct charges was likely intended to maximize reputational damage to the Sasisekharan Lab.
General and Scientific Rebuttal:
The following facts prove that the allegations in the mAbs perspective against the Sasisekharan Lab at MIT were baseless.
- A detailed comparison of the engagement of ZAb_FLEP antibody with the FLEP epitope on the Zika virus (ZIKV) based on a cryo-EM structure and EDE I C8 antibody with the EDE epitope was peer-reviewed and published in 2020. This study clearly articulates the distinct engagement of ZAb_FLEP with the FLEP epitope that is different from the EDE I C8 antibody with the EDE epitope despite the overlap between these epitope regions on the surface of the ZIKV.
- An independent validation of the Sasisekharan Lab’s antibody engineering platform and the details of the implementation of this platform to engineer ZAb_FLEP was published in Antiviral Research in 2021 (referred to as AVR paper). This peer-reviewed study clearly shows that ZAb_FLEP was not engineered by making random mutations on EDE I C8 antibody as alleged by the mAbs opinion paper.
- VIS410 is an anti-influenza antibody that was designed, created, and invented by Visterra Inc., and at Visterra Inc. Also, Visterra Inc., was granted a patent for its novelty by the US Patent office in 2014 (US8877200B2). The inventors on this patent included only Visterra’s employees and did not include Prof. Sasisekharan or any of Sasisekharan Lab members from MIT.
We will focus on the science regarding ZAb_FLEP and its connection with C8 because it was developed in the Sasisekharan Lab in conjunction with the Sasisekharan Lab’s SMART-related activities in Singapore.
There are several recurring implied allegations in the mAbs perspective which can be comprehensively addressed in the following.
Strawman Argument that Sasisekharan Lab claimed an unsupervised in silico algorithm was used to create de novo ZAb_FLEP
One of the baseless and ludicrous premises of the mAbs article is built on the strawman argument that:
- Sasisekharan Lab used in silico algorithms (which are typically programs that take in an input and deliver an output without any human intervention or are “unsupervised”) to design the antibodies.
- In silico de novo design methods have been unsuccessful in designing antibodies and so it follows that Sasisekharan Lab falsified the methods.
The Sasisekharan Lab has never stated or claimed anywhere or in any publication that the Sasisekharan Lab antibody-engineering platform involves de novo design with exhaustive in silico engineering. Instead, Sasisekharan Lab approach relies on available information obtained from distinct antibody parts, which are assembled and then engineered using key metrics that the Lab developed over the past decade and published in many peer-reviewed top tier journals (PMC2826414, PMC3242012, PMC3650665, PMC4258880, PMC5387726, PMC3637786, PMC4758460, PMC5981469).
Had the Adimab authors read the online pages of the CHM paper, they would have found on page e2, and running through page e6 was a summary of the methods used to develop and test ZAb_FLEP. Computational tools are mentioned frequently, and seven are listed in the Key Resources Table. Only two are from the Sasisekharan Lab, Multivariate Logistic Regression (MLR) and Amino acid Interface Fitness (AIF), and they are described in detail in a 2013 PNAS paper. All the rest are publicly available, and none could be characterized as an “unsupervised algorithm.” The CHM paper summary is more easily understood if one examines the Sasisekharan Lab website and clicks through to Antibody Engineering Methods. There the Adimab authors would have found a detailed explanation of the antibody engineering methods that had been developed in the Sasisekharan Lab and that it includes significant intervention by human expert during the process of target epitope identification, template selection, scaffold screening and final design candidates ranking.
The following animation provides an overview of the 4 steps.
Allegation: Misleading origins of the VIS_410 and ZAb_FLEP antibodies due to high sequence similarity with FI6v3 and EDE I C8 respectively and therefore VIS410 and ZAb_FLEP are plagiarized or disguised versions of FI6v3 and EDE I C8 respectively
This linear sequence identity argument is absurd for several reasons. First, the linear sequence identity in the heavy and light chains within a given human IgG subtype is already >90% and can approach 95% for post- Phase 1 biologics. Therefore, using a linear sequence identity argument is incongruous.
The limitations of such sequence analysis were dramatically demonstrated in late November 2021 when Adagio Therapeutics asserted that their antibody ADG-20 would neutralize the Omicron variant of SARS-COV-2. They came to this conclusion based on sequence similarity between the Omicron variant and the past variants of the epitope surface targeted by ADG-20. By mid-December it was reported that ADG-20 took a 300-fold hit in neutralization potency against Omicron. This validated the predictions from the Sasisekharan Lab, released a week earlier, that epitope surface changes on Omicron would reduce the potency of many SARS-COV-2 antibodies including ADG-20.
The US PTO and several other agencies recognized the novel and nonobvious differences in sequences between VIS410 and FI6v3 and between ZAb_FLEP and EDE1 C8. Moreover, such misleading linear comparisons neglect a key point: the same VH and VL sequences inserted in different antibody frameworks (FR) can have dramatically different antigen recognition properties. The FR and complementarity determining regions or CDR (combined VH and VL) work together to define the overall antibody engagement with the target epitope in the three-dimensional structural space.
Had the Adimab authors read the online pages of the CHM paper, they would have found on page e1 the Key Resources Table. Listed in that they would have found “Human recombinant C8” from the Protein Data Bank, with the identifier, “RRID: AB_2725799;PDB: 5LBS.” Such a listing is not the tool of a plagiarist.
As the CHM paper made clear, C8 was one of the templates used to build ZAb_FLEP. The role of C8 as explained in the CHM paper is summarized in the context of our Antibody Engineering Methods following:
- C8 first appeared in the selection of templates, after the Epitope had been selected.
- C8 failed in scaffold selection and was NOT selected as a starting scaffold.
- C8 appeared again in paratope engineering as it was carried over as a Template Candidate. C8 was selected as a CDR template, and then was modified to optimize its binding to the identified Target Epitope.
This last point explains why the CDR of ZAb_FLEP is similar to C8, while having important functional differences The most conclusive evidence to quash this implied allegation is presented in an independent work published in Antiviral Research by members of the Sasisekharan Lab who were not involved in the 2018 CHM paper and employees of Tychan Pte Ltd – a company founded by Prof. Sasisekharan in Singapore. This work shows a detailed independent implementation of the methods described in the CHM paper to arrive at ZAb_FLEP and highlights the functional differences between ZAb_FLEP and EDE I C8 and the engineering steps that led to these differences. This work clearly establishes: 1) there was no reason for the Sasisekharan Lab to hide the sequence or the details of ZAb_FLEP engineering; and 2) the role of the relationship between CDR and FR of antibodies to explain the nuances of the difference between antibodies which cannot be inferred from sequence similarity. As an example, in the AVR study, when the EDE I C8 CDRs were inserted into the FR region of ZAb_FLEP, these EDE I C8 CDRs no longer recognized the antigen despite this sequence sharing more than 90% sequence similarity to EDE I C8.
Allegation: Only conservative and inconsequential mutations were randomly made to CDRs of the alleged antibodies of origin viz. FI6v3 and EDE I C8 to generate VIS410 and ZAb_FLEP respectively.
ZAb_FLEP antibody was not created by introducing functionally inconsequential mutations into the EDE1C8 antibody. Instead, ZAb_FLEP was engineered with a design criteria in mind by combining VH, VL (scaffold) and CDR features (templates) from three different antibody sources to target a distinct quaternary epitope surface: VH scaffold of antibody AL-57 targeting human integrin LFA-1 (obtained from PDB: 3HI5 chain H), VL scaffold of anti-CMV antibody (obtained from PDB: 3EYF chain A) and select CDRs (HCDR2, HCDR3, LCDR1, LCDR3) of C8. ZAb_FLEP is different from C8 both structurally and functionally. ZAb_FLEP targets a distinct quaternary epitope not targeted by any known antibody in the field. Secondly, ZAb_FLEP shows poor neutralization of dengue serotypes 3 and 4 consistent with its specificity towards ZIKV; on the other hand, C8 neutralizes all four serotypes potently. As mentioned above, these results were also validated in the AVR study where a side-by-side comparison of ZAb_FLEP and C8 was performed (AVR Appendix I). Ser and Tyr along with Gly and Arg are the most abundant amino acids in naturally occurring antibodies’ CDR loops (also seen in Sasisekharan Lab AIF matrix). In fact, antibody libraries created by randomizing CDR sites with only these four amino acids have yielded functional antibodies. Because of their ability to mediate most of the contacts necessary for high-affinity antigen recognition, substitutions to Ser (from Thr) and Tyr (from Phe) become obvious design choices. The “non-conservative framework mutations” are a signature of the starting VH and VL scaffolds that were used for the engineering of ZAb_FLEP.
Summary
- The methods employed in the CHM paper were developed by the Sasisekharan lab over the last 10+ years based on several peer-reviewed papers published in top tier journals.
- Contrary to the mAbs perspective’s misrepresentations and biased opinion, no unsupervised algorithms were used in the study. The origins of the CDR mutations and framework diversity have been incorrectly captured and described in the perspective.
- The mAbs perspective takes the myopic view that only CDRs are important for function whereas the rest (e.g., FR) are largely irrelevant, which further underscores this narrow view of antibody structure-function.
- The computational process leading from epitope selection to design of ZAb_FLEP was independently recreated by other researchers.
- The very public nature of method misrepresentation, media leaks, commercial and other conflict of interests together with formal allegations of research misconduct point to a malicious intent.
