Good day, Shamik Mitra, PhD. I'm Arjun. I've had a look through your background and your professional record. I'm interested in hearing more about your work and how you see yourself fitting in here with our faculty. Ready to start?

Please start by introducing your professional journey in academia.

So I have done my PhD in bioinformatics and cancer genomics. After that, I worked as a postdoctoral researcher at Lund University under Professor Frederick Martins.

Worked with exploring the molecular characteristics of the soft tissue sarcomas.

And.

Thank

Then I.

Left academia and I joined.

And there I worked with making the CRISPR CAS 9 based genome editing safer with developing the safety module.

CRISPR CAS 9 gene editing startup called Helix.

And there I worked. I laid the transcriptomic analysis of head and neck commercial carcinoma tumors from the patient samples undergoing a novel immunotherapy.

After that I joined Rakuten.

And there I and also I also published 1.

Poster.

In a conference.

With AS 1/4. After that I joined Accelerant and in Exceldra I worked on developing reproducible workflow, bioinformance workflow in the platform.

That we developed for a custom platform as a product and there I work with the SAS environment for past one year.

Thank you for providing more insights into your professional journey. Let’s delve deeper into your expertise and achievements to better evaluate your fit for this role. Could you elaborate on the specific challenges and strategies you employed during your transition from academia to industry, particularly at Helix and subsequently in your roles requiring transcriptomic and bioinformatics expertise?

Yes. So one major difference with academia and industry that I faced is that industry is more of a breadth first than depth first. Academia is more about delving into specific problems and trying to understand the problem with deeper meaning, whereas the industry it is more about solving the problem. So basically like let's say you have like.

A problem X so in academia what we'll try is to understand that what this problem makes entails does it have any association with another problem why Zed whatever and then we go into finding like what could be the plausible solutions for problem X whereas in industry.

The main idea is like what is problem X? How do we define the problem and then we go directly into the solution we.

Look for the dependencies much less. So this is one challenge and the other challenge is that academia you have much more flexibility about time. So you can spend a lot of time exploring things. Whereas in industry the exploration is always fast. You do this spikes or investigations really fast within like a week or two weeks at most and then you go into directly solving the problem.

So it is more about a problem solving approach at the industry with a fast iteration.

That distinction between academia's depth-oriented research and industry's problem-solving focus with rapid iteration is insightful. In your transition, how did you adapt your academic training in deep, foundational analysis to fit the faster cycles and solution-driven demands of your roles at Helix, Rakuten, and Accelerant? Could you share a specific example?

Yes. So basically, for example, when I started with the problem in Helix about the safety module, the main challenge was to understand that what exactly would entail in the safety module, for example, whether we are. Let's give you a concrete example. So basically we were dealing with a genome editing guide RNA and this guide RNA would bind to a certain position in genome. So now we have to see that whether.

...this guide RNA binds exclusively to the intended target or if it has potential off-target effects that could compromise safety. How did you approach such optimization, and what steps or tools were critical in ensuring that the safety module you developed would reduce risk while maintaining the efficacy of the CRISPR Cas9 system?

Hello.

Yes, Professor. Could you continue explaining how you managed the challenge at Helix in designing the safety module, specifically handling issues like off-target effects with your gene-editing approach?

Yes. So basically what I worked on is to understand that what kind of mutations are there in the region of interest and then like if there are sequence homologies and most important factor is the chromosomal translocations. So if the region that is cutting is prone to chromosomal translocations, it would cause downstream carcinogenic effects. So all these things including keeping in mind.

The 3D structure of the genome which is also very important for the chromatin assembly. So I worked on maintaining all these factors and develop AIML guided scoring system which would ensure that it will score the guides with different scoring points based on all these measures and then we will select the guide which is safer.

But in this coding system.

You mentioned developing an AI/ML-guided scoring system to evaluate CRISPR guide safety based on factors like mutation types, sequence homology, chromosomal translocation risks, and the 3D genome structure. Could you clarify the types of models or algorithms you utilized to integrate these diverse factors into a unified scoring system? Additionally, how did you validate the effectiveness of this system?

Yes.

So this guide, uh, said this guide selection was motivated by several different factors such as free energy, the.

Sequence homology and the sequence itself with the GC content and all these different factors were incorporated into a convolutional neural network model where the outcome was basically the editing percentage.

And based on that we trained the model with a multi layered convolutional neural network and then we understood that which guides are based on that we selected the guides that are doing better.

Understood, Professor. Leveraging a multi-layered convolutional neural network to predict editing efficiency while incorporating various factors like free energy, GC content, and homology demonstrates an advanced approach. Could you share how you ensured the robustness of your model, such as the dataset size, validation techniques, or experimental confirmation, to guarantee reliability for real-world applications?

Basically we gather datasets.

Both train and test data from publicly available data and based on that data set we used, we partitioned the data into training and then.

CV or cross validation and then finally testing and based on the test results we validated the model.

Thank you for the explanation, Professor. That structured approach of using publicly available datasets with training, cross-validation, and testing to validate your model's performance illustrates thoroughness. Shall we now explore your contributions to transcriptomic analysis during your tenure at Rakuten?

Absolutely.

So in.

Yeah, in rec.

Take your time, Professor. Could you delve into your role in leading the transcriptomic analysis of head and neck carcinoma tumors at Rakuten, focusing on the methodologies or tools you employed and the key insights you uncovered?

So basically the problem that I faced in Rakuten was about.

Different types. So basically it was about novel immunotherapy called photo immunotherapy. And in this photo immunotherapy, there were eight patients that undergone the immunotherapy and we wanted to understand like how they respond. I mean we had temporal study with before and after the treatment and we wanted to understand like how the patients respond.

Based on the immunotherapy with the different immune cell subsets that are present in the tumor.

And for that, the study was based on single cell RNA C transcriptomic analysis and I basic I.

This was basically made by 10X Genomics and I used their 10X proprietary technology Cell Ranger to process the data. And then I brought then I worked with the process datasets in Surat to uncover the different immune cell subsets that are present in the tumour, their percentages. And then we wanted to understand like how this.

Differs between the patients before and after immunotherapy.

Evoked a lot of neutrophilic response in the tumors because of a lot of necrosis happening due to photo immunotherapy and subsequently we also found some other immune cell subsets of importance and.

And based on that, we understood that it it basically it it.

Based on that it the study was concluded.

A fascinating study, Professor. Conducting single-cell RNA sequencing to analyze immune cell subsets and using tools like 10X Genomics' Cell Ranger and Seurat illustrates technical depth. Could you elaborate on how these findings, such as the neutrophilic response and immune cell subsets, shaped further discussions or potential improvements in the phototherapy approach?

Yes, the neutrophilic response is very common because usually after someone undergoes immunotherapy and especially photo immunotherapy that causes a lot of necrosis in the tumor because of the high intensity photo beam.

What happens due to this necrosis that usually the immune response can be divided into two parts, the innate immune response and the adaptive immune response. Since these samples were taken just 24 hours after the photo immunotherapy, so there was no chance of any adaptive immune response that we will be able to capture. So it was mostly based on the inept immune response and inert immune response. Neutrophils are very common in the tumor microenvironment and they have both the pro and anti-tumulogenic properties.

However, in these cases, these neutrophils.

They indicated a good response for the patients as we gathered from the later patient data and based on that these neutrophils.

These convert the patients with suitable clinical benefit.

We also found some other immune cell subsets, for example like type of regulatory T cell.

Which also can block the tumor immune response and can be of tumor associated inflation, inflammation. So overall what we found that these neutrophilic response is a good indicator of the patient prognosis undergoing photo immunotherapy immediately after the photo immunotherapy session.

Thank you, Professor, for detailing that. Your work in correlating neutrophilic responses and regulatory T-cell dynamics to patient prognosis showcases a nuanced understanding of tumor immunology. Shall we now focus on your bioinformatics platform development at Accelerant?

So in Excel rack, the kind of problem I face, it is more of an engineering problem rather than a research problem. So we had a lot of different.

Sure.

Bioinformatics pipelines processing the NGS data, whether it be chipset, whether it be bulk harnessing single cell RNA seq.

Genomics, for example, like mutation from whole genome or whole exome sequencing, both somatic and John line variant calling. So all these pipelines, we modified them and we process them to be incorporated into a platform where it can be run by.

And uh.

Biologists not buying from editions with very limited knowledge of bioinformatics. So it is something that is very user friendly and very easy to access. And then we also provided downstream analysis such as like.

EU maps for single cell RNSC or for bulk RNC, the different differential gene expression analysis as well as the volcano plots and everything to understand how the expression works as well as genomic pathway annotations using.

The tools such as GO terms analysis and Reactome pathway analysis.

That's an innovative solution, Professor. Transitioning bioinformatics pipelines to a user-friendly platform for biologists not well-versed in computational techniques is a significant contribution to widening access to these tools. Could you clarify how you ensured the scalability, reproducibility, and user accessibility of this platform for such a diverse dataset?

Yes. So for example, for this particular platform, it was hosted in AWS.

And in AWS, we use the AWS batch to process this datasets fast so that they can be processed in parallel with different processes running in tandem. And then the reproducibility we tested with different datasets and in these datasets we every data set for every organ, we, we support it for certain organisms and each Organism we have datasets and we tested them, we validated the results, we made case study out of it and.

On basis of that, also like we took studies that were available in public repository. So basically we can validate whatever results we found against the the original papers or articles, scientific articles that shown these results and we validated them. We found validated the findings and that's how we ensure that it's a reproducible and as well as we had run multiple times every data set and we ensured that we get the same result.

So that way we ensured that and as well as we tried with different sizes of datasets and we saw that the datasets scale in terms of the sizes and the timing of the datasets, the number of samples and everything and each sample wise we calculated costs as well.

No surprises.

Understood, Professor. Hosting on AWS and leveraging AWS Batch for parallel processing, combined with rigorous validation and cost analysis, clearly exemplifies solid engineering and scientific diligence. At this point, I believe we have comprehensively explored various aspects of your expertise. Would you like to add anything, or shall we conclude the interview?

UH-1 experience that I did not talk in detail is about my post-op and in my post doc I worked with so I was the only band from T Shin and there I worked with every sort of data set like single, not single cell. Sorry the bulk currently seek the gene fusion analysis.

As well as a lot of whole genome and whole exome analysis. And there I developed the pipelines as well, like I did here. And I use these pipelines to explore the data. And then also with the transcriptomic analysis, I worked on different projects to understand sarcoma, different types of sarcoma better and what. Constitutes different subtypes and their prognosis as well.

The studies I have published I've been part of four publications, 2 of where I was the first author, one I was the co-author and the other one I was a shared last author.

And based on that.

Thank you for adding that, Professor. Your postdoctoral work reflects a deep immersion in diverse datasets and complex analyses, alongside contributions to foundational research through multiple publications. If there’s nothing further you’d like to cover, would you like to confirm that we can conclude the interview?

If you have no further questions.

Yes, we can.

Yes.

Thank you, Professor. I believe we’ve thoroughly covered the relevant aspects of your experience and expertise. Would you kindly confirm once more if you'd like to officially conclude the interview?