Antibody humanization has reached certain technical maturity; nevertheless, there are plenty of spieces on earth in need of antibody drugs that lack a
proper solution. We aim to tackle the issue with artificial intelligence for its powerful fitting, learning, and predictive capabilities. On one hand,
we build an extensive homologous library of antibody FRs (Framework Regions) and CDRs (Complementarity Determining Regions) from various species. On the other hand,
we optimize antibody immunogenicity via mutagenesis of a given sequence and adapt multi-dimensional scoring tools.
2. Proposing a rational approach for scoring antibody immunogenicity, encompassing diverse species and suitable for species with limited existing data.
3. Delivering a comprehensive set of automated tools for designing antibody for diverse species, enabling multi-dimensional evaluation of sequences.
Since establishing contact with VJTbio in July, <strongstyle="color: #8a61ad;">we have engaged in multiple friendly exchanges and discussions, establishing a close contact.</strong>
Our communication with VJTbio has been continuous <strongstyle="color: #8a61ad;">throughout the entire project</strong>, and at each stage of the project, our collaboration with
the company has provided us with significant inspiration. It has helped us clarify our direction and insist on our original intention of
creating an iGEM project that is meaningful to the industry and the world.
</p></div>
<h1class="title"id="Header2">Background</h1>
<divclass="fade-in"><p>
Antibodies are proteins produced by the immune system in response to the presence of foreign substances called antigens. Their primary function is to recognize
and neutralize these antigens to protect the body from infections and diseases. Moreover, the antibodies obtained in one species (e.g., murine) cannot be directly
used in another species, as it may lead to the development of an immunogenic antidrug antibody (ADA) response.Therefore, to obtain antibodies for a specific species
without conducting in vivo experiments, it is necessary to perform a process called <strongstyle="color: #8a61ad;">"species-specification", designing antibodies suitable for certain speices basing
on those of other speices, on existing antibodies such as mouse-derived antibodies.</strong> This process involves modifying the antibodies to reduce their immunogenicity when
On July 18, 2023, despite still being in the summer vacation period, team members who remained in Beijing
formed an advance detachment and departed with anticipation for VJTbio. During the exchange, we gained
a clearer understanding of the great significance of producing animal antibodies and realized <strongstyle="color: #8a61ad;">the substantial
disparity between the technical level of animal antibodies and that of human antibodies</strong>; therefore providing
us with ample room for exploration.
</p>
<p>
VJTbio primarily validates the properties of antibodies through wet lab experiments. Thus, during the communication, they also
obtained insights into the progress in the field of AI from us, and they are eagerly looking forward to incorporating AI in
various aspects of their production process.
</p>
<p>
Our team had been conducting research on AI design for "humanization of murine antibodies". Through our initial communication with VJXbio,
we learned about their achievements in the field of "antibody caninization" and the significant demand in the field of animal antibodies.
We aimed to <strongstyle="color: #8a61ad;">broaden our original Topic, which is "Humanization of murine antibodies" to encompass various "species-specification" of antibody</strong>, and the research we previously conducted
on humanization of murine antibodies could still be applicable in this context. After an intense week of thorough literature review and
preliminary technical validation, we ultimately settled on a viable technical solution.
</p>
</div>
</div>
<h4class="title"style="color: #573674;">online communication </h4>
<divclass="fade-in"><p>
We established WeChat groups with key members of the company, frequently seeking advice, sharing insights, and learning from one another within the group.
Online communication is highly convenient, and we have also received warm and friendly treatment from the company. We are deeply grateful for the patient
guidance and numerous inspirations provided by company.
</p>
<p>
In our early research, we found out that the field of humanization of murine antibodies has developed since the 1980s and has now achieved considerable biotechnological
maturity. Nowadays (especially since 2023), significant breakthroughs in antibody design have been reached with the help of AI technologies, especially in the highly
variable CDR H3 regions. In this context, methods such as transgenic mice, antibody humanization, and AI-driven de novo design are all proven effective in obtaining antibodies
suitable for human use. That is, the realm has come a long way in designing antibodies with high affinity and low immunogenicity to human.
We have had discussions regarding the <strongstyle="color: #8a61ad;">reliability</strong> of antibody de novo design and the <strongstyle="color: #8a61ad;">recognition of antibody de novo design within the industry</strong>.
We have come to realize that the application of antibody de novo design is not yet mature, and there is also a lack of recognition for de novo
design within the industry. AI companies often obtain a large number of positive antibody sequences as a training dataset and then use this as a
basis for prediction to design antibodies with higher affinity. So we also chose <strongstyle="color: #8a61ad;">extending the spieces derived from murine antibodies with AI
methods in the project instead of de novo design. </strong>
</p>
<p>
However, after communicating with a biotech company, we realized that there are still plenty of gaps left to fill in designing antibodies for various species other than human.
Regarding <strongstyle="color: #8a61ad;">the verification of our project</strong>, we also consulted the company about the verification method in the industry.
The company introduced a mainstream method to us, which is to express and purify the full-length antibody sequence including the designed FR and CDR, and then
perform Protein Protein Interaction Experiments with antigen protein. Through such communication, we decided to add <strongstyle="color: #8a61ad;">structure score of the antibody sequence</strong>
to our project, which is divided into two parts: on the one hand, whether the antibody sequence can be folded into a normal structure; on the other hand, we tend
to maintain the original structure to <strongstyle="color: #8a61ad;">retain antigen-antibody binding</strong>.
</p>
<p>
For example, antibody drugs are required for whether pet dogs, pet cats, or animals like pigs, cows, and sheep that have significant importance in the livestock industry.
Therefore, attempts to conquer the problem are neccessary, and our goal is to extend the progress made in antibody design to a broader range of species with the help of artificial intelligence.
We are very thankful for the patient guidance and numerous inspirations provided by company.
</p></div>
<h1class="title"id="Header3">Our Solution</h1>
<divclass="fade-in"><p>
During our discussion with the company, they offered us many insights and advice to the realm and industry. Consequently, between the two technical roadmap, AI de novo design and
<strongstyle="color: #8a61ad;">extending the spieces derived from murine antibodies with AI methods</strong>, we chose the latter one as the former technique hasn't been widely acknowledged by the industry yet and is unable to achieve industrial-level reliability.
As students with an <strongstyle="color: #8a61ad;">information background</strong>, we believe that it's necessary for us to enhance
our understanding of synthetic biology and gain insights into the industry by conducting <strongstyle="color: #8a61ad;">on-site visits.</strong>
This will further improve our iGEM project.
</p>
<p>
Antibodies are divided into constant and variable regions, within the variable region are the high-frequency mutated CDR regions (where the antigen comes into contact with the antibody), and the relatively conserved and species-specific FR regions.
In addition to visiting VJTbio, we have also visited numerous biotechnology-related companies throughout
the season, gaining valuable experiences from these field trips.
</p>
<p>
On this note, our project is divided into two main parts.
During the visit and exchange at QiTan Tech, we not only toured many of the company's bioinstrumentation, but also gained
a deep understanding of the structure, properties, and meaningful applications of nanopore proteins. The company's years of efforts
in optimizing nanopore proteins are precisely what have led to the company's current level of accuracy in genetic sequencing.
</p>
<divclass="fade-in">
<p>
We engaged in thorough discussions concerning protein optimization and design. During these discussions, we conveyed our team's
interests and the directions we intend to explore. Also，we learned that the company's algorithms in designing and optimizing nanopore
proteins have room for improvement, and our team believed that there is potential for collaboration in this area with the company.
However, through further dialogue, a consensus was reached between our team and the company. Namely, the company was focusing on
enhancing sequencing accuracy, and even incremental progress from 99% to 99.5% holds significant importance within the industry.
Nonetheless, this endeavor demands substantial time for accumulation and rigorous validation. As undergraduate students, rather than
graduate students, we probably should embark on tasks that leverage the creativity and imagination more inherent to undergraduates.
</p>
<p>
Furthermore, we also engaged in discussions with the company about the breakthroughs in the field of de novo protein design in 2023.
During the exchange, we learned that <strongstyle="color: #8a61ad;"> "de novo protein design" is highly challenging</strong>, currently achievable mainly by the David Baker
research group. This realization subsequently led us to consider focusing on antibodies with more predictable sequences and structures,
and <strongstyle="color: #8a61ad;">continuing our efforts in the design and optimization of antibodies</strong>. It was after this that we began to pay attention to learning many aspects of antibodies,
including the antibody structure prediction tool IgFold, the progress of de novo design of antibody CDR-H3 loops, antibody language models, etc.
</p></div>
<h3id="ZGC"style="font-weight: bold;color:#573674">Zhongguancun Life Science Park in Beijing</h3>
<divclass="fade-in"><p>
We specifically visited the Zhongguancun Life Science Park, which focuses on cutting-edge tracks such as cell and gene therapy,
innovative pharmaceuticals, AI-driven drug development, AI-assisted diagnostics, and more. We visited some companies within the park,
with a particular emphasis on holding in discussions with Beijing Syngentech and BeiCell Biotechnology.
</p>
<divclass="row"style="margin-top: 0rem;">
<divclass="col-lg-9"style="margin-top: 0rem;">
<h4id="ZGC"style="font-weight: bold;color:#573674">Syngentech (On May 28th)</h4>
<p>
<strongstyle="color: #8a61ad;">Company profile:</strong> Beijing Syngentech is dedicated to the research and development of gene
and cell therapy drugs based on synthetic biology technology, along with providing scientific research and clinical services in this field.
Through engagements with both companies, we not only toured many of the company's biotech instruments, enriching our biological background knowledge,
but also gained a deeper understanding of <strongstyle="color: #8a61ad;">the role of synthetic biology in biomedicine and life health</strong>. We learned that AI has already played a
significant role in new drug development. What's more, both companies have utilized AI to optimize wet lab processes, making substantial contributions
to clinical diagnosis, treatment, and drug development.
</p>
<p>
The first part is collecting antibody variable region sequences from various species and using them to construct a homologous library of FR (Framework Region) and CDR (Complementarity Determining Region) for each species, providing
a comprehensive and reusable resource for future research.
After this visit, we have all developed a strong interest in the role of AI in facilitating new drug development. Through our further research,
we found that <strongstyle="color: #8a61ad;">antibody drugs constitute a significant portion of novel pharmaceuticals. </strong>We hope our project can contribute to the optimization
and design of antibody drugs and ultimately established a close collaboration with VJTbio as previously described.
</p>
<p>
The second part highlights antibody immunogenicity optimization based on mutations to acheive our goal of antibody species-specification. Meaning, it centers on antibody FR + CDR design, drawing from the achievements in the field of antibody humanization to avoid potential immunogenic
antidrug antibody (ADA) response when applying antibodies to a new species.
Additionally, during the exchange, we also acknowledged the critical significance of "safety" in the new drug development process. In our project,
we placed significant emphasis on incorporating measures to <strongstyle="color: #8a61ad;">ensure the security of data sources and the integrity of algorithmic output sequences</strong>.
We hold the view that comprehending how AI models predict and optimize antibody sequences and characteristics is crucial for ensuring their reliability and safety, so we pay more attention to <strongstyle="color: #8a61ad;">develop more interpretable codes and algorithms of AI models</strong> in the process of designing and optimizing antibodies.
</p></div>
<h1class="title"id="Header4">What We Are Doing In Detail</h1>
<divclass="fade-in"><p>
In detail, in the first part of the project, due to the lack of PDB files of many antibodies from different species, we first expand the dataset via BLAST to search for homologous sequences and valuable germline gene framework,
<h5style="width:50%;font-weight: bold;margin-bottom: 12px;">(Chinese Academy of Sciences, Assoc. Prof.)</h5>
</div>
<p>
On April 19th, we had the privilege of inviting Yinglu Cui from the research group of Professor Bian Wu at the Chinese Academy of Sciences to visit Tsinghua University. We held a group
seminar where we engaged in extensive discussions regarding protein design and enzyme engineering.
Professor Cui introduced to us the research undertaken by their group in the field of "computational-enzyme design." We reviewed the history of protein optimization and design,
highlighting areas such as rational design, semi-rational design, directed evolution, etc. She outlined the <strongstyle="color: #8a61ad;">various breakthroughs and ongoing challenges in the transition to the
current phase of computational design</strong>. Further, she provided insights into the prospects of applying computer-designed non-natural enzymes in the field of biomedicine.
</p>
<p>
In the second part, considering there are fewer data for antibody of different species in comparison to human antibodies, we propose an immunogenicity optimization-based strategy. Conventionally, antibody humanization approach involves
CDR grafting and back mutations, and the initial step of such method entails aligning homologous sequences from a homologous library to serve as a template for further design. Hence, the approach may not be applicable to species with
small datasets, as the results of the homology alignment may lack reliability due to the limited number of samples. Thus, in our project, we identify the CDRs and FRs of antibody sequences and guide the directed evolution through
mutagenesis of given sequences using immunogenicity scoring. In essence, our algorithm intend to adapt existing antibody humanization scoring tools to be applicable for other species with limited data. Afterwards, we apply other
scoring methods on mutiple scales such as structure score and affinity score to given antibodies.
Our team presented our research and algorithm designs in protein optimization, sharing some of the algorithmic models we've experimented with. Professor Cui shared valuable insights
from her lab's experience in this field, pointing out that the one-hot encoding we've been using lacks universality. Instead, she suggested we should include <strongstyle="color: #8a61ad;">pre-encoding work </strong>and emphasized
the importance of assessing the <strongstyle="color: #8a61ad;">generalization capability</strong> when researching model algorithms in the protein design realm. She noted that some current models, such as DLKcat, suffer from
weak generalization abilities. In response to Professor Cui's suggestions, we systematically worked on improvements, addressing each point. Especially in terms of pre-encoding work, we have done a lot of pre-encoding research and practical attempts.
Through constant comparison and trade-offs, we finally chose <strongstyle="color: #8a61ad;">BLOSUM62 </strong>in our project, which is the most suitable for our project, <strongstyle="color: #8a61ad;">taking into account a large number of rare species.</strong>
</p></div>
<h1class="title"id="Header5">Why AI</h1>
<divclass="fade-in"><p>
In this era of big data, we can mine many potential sequence rules from existing sequences. Owing to the potent fitting, learning, and predictive capabilities of AI algorithms, applying AI solutions to biotechnological problems offers numerous
benefits, including accelerating research and discovery, personalization, enhanced data analysis, and guided antibody engineering decisions, etc. AI algorithms can create a cross-species antibody design model by analyzing antibody
sequences and structural data from different species. The model can extend antibody design to more species, including pet dogs, pet cats, and livestock animals. By analyzing vast amounts of antibody sequences and structural data,
AI can identify crucial sequence and structural features, providing valuable guidance for antibody engineering decisions.
<strongstyle="color: #8a61ad;">The role:</strong> Professor Xie has extensive experience in both academia and industry, providing us with guidance on the
direction of our research topics and technical approach.
</p>
<p>
<strongstyle="color: #8a61ad;">Profile:</strong> Professor Xie Zhen is a professor at the Center for Synthetic and Systems Biology at Tsinghua University,
as well as the founder and chief scientist of Syngentech.
The integration of AI's automation technology enables efficient screening and discovery of new potential antibody candidates from extensive antibody libraries, significantly increasing the efficiency and success rate of antibody discovery.
Furthermore, AI algorithms automate the simulation and computation of numerous antibody sequences to optimize their properties such as affinity, specificity, stability, and production efficiency. This contributes to enhancing the therapeutic
effectiveness and production efficiency of antibodies.
After communicating with several protein design and optimization companies, we have transitioned our focus from protein optimization design to antibody optimization design. However,
we haven't yet identified a specific task. Before establishing a concrete plan, we learned about recent breakthroughs in using artificial intelligence for de novo antibody design and
extensively reviewed relevant literature. Then, we discussed the details with Professor Xie. After listening to our existing research and algorithm attempts, as well as our interest in antibody drugs,
he pointed out a meaningful background for us: <strongstyle="color: #8a61ad;">humanization of murine antibodies</strong>. We involved in thorough discussions about the context, significance, and implications of such theme.
</p>
<p>
Our AI approach brings several advancements in antibody species diversification. First, with computational-guided directed evolution and rational analysis of antibody sequences, our project can markedly reduce the material resources and time
consumed in wet-lab experimental validations. Secondly, we gather large scale data in our homologous library of FRs and CDRs in diverse species in an organized manner, which provides added convenience for further antibody-related reseach.
Last but not least, in terms of drug development, with the comprehensive knowledge of given antibody sequences generated by our algorithm and providing antibodies suitable for multiple species, we can expedites the drug and treatment
development process and open up new possibilities for disease therapies.
We further investigated antibody humanization and found that traditional humanization techniques primarily involve a series of biological tools and simulation computations implemented
on computers, whereas machine learning deep learning methods are less used in the field, and are mainly employed for binary classification and humanization scoring of humanized antibodies.
We aim to <strongstyle="color: #8a61ad;">delve deeper into using machine learning and deep learning tools to address challenges in this domain, incorporating considerations for antibody structures. </strong>When we propose a preliminary
technical route, we once again communicated with Professor Xie, who acknowledged our idea and expressed anticipation for us to construct a comprehensive automated tool and add our innovative ideas.
<strongstyle="color: #8a61ad;">The role:</strong> Ziting Zhang provided us with an in-depth introduction to immunological knowledge related to antibodies. As a researcher in the field of immunology,
she expressed that our project holds significant meaning.
</p>
<p>
<strongstyle="color: #8a61ad;">Profile:</strong> Ziting Zhang is a Ph.D. student in Bioinformatics at Tsinghua University, and she possesses unique insights in immunology.
Moreover, Zhang Ziting was a member of the Tsinghua-A team in iGEM 2020.
Ziting Zhang also pursued her undergraduate studies in the Department of Automation at Tsinghua University. Currently, her research interests converge at the intersection of bioinformatics. As a scholar in immunology, she believes that our team's project this year holds considerable meaning.
Through exchanges with her, we gained a deeper understanding of immunological knowledge</strong> related to antibodies and <strongstyle="color: #8a61ad;">how to leverage the advantages of our team's background in information. </strong>
Through this exchange, we spent more energy in the project to improve our algorithm. On the one hand, due to our proficiency in algorithms and rich experience in tuning parameters, we have extensively tested many machine learning and deep learning models,
and finally selected the most reasonable one by comparing the <strongstyle="color: #8a61ad;">performance</strong> and <strongstyle="color: #8a61ad;">mathematical principles </strong> of each algorithm. On the other hand, through our background knowledge and algorithm experience, we have done a lot of <strongstyle="color: #8a61ad;">fine-tuning</strong> to make the whole model more perfect.
</p></div>
<h1class="title"id="Header6">References</h1>
<divclass="fade-in"><p>
[1]Claire Marks and others, Humanization of antibodies using a machine learning approach on large-scale repertoire data[J/OL], Bioinformatics, Volume 37, Issue 22, November 2021, Pages 4041–4047.
<strongstyle="color: #8a61ad;">The role:</strong> Xiangzhe Kong provided us with a lot of advice on CDR algorithm design, which helped us to understand the current situation and cutting-edge progress
in the field of antibody CDR design.
</p>
<p>
<strongstyle="color: #8a61ad;">Profile:</strong> Xiangzhe Kong is a Ph.D. student at the Department of Computer Science, Tsinghua University, and the author of "Conditional Antibody Design as 3D Equivariant Graph Translation"
(ICLR 2023 Outstanding Paper Honorable Mention). He possesses extensive experience in computational de novo antibody design, especially CDR design.
Antibodies are divided into constant and variable regions, within the variable region are the high-frequency mutated CDR regions (where the antigen comes into contact with the antibody), and the relatively conserved and
species-specific FR regions. As the antigen-antibody binding region, CDR has garnered significant attention in the academic field, and there are a lot of difficulties in the design of CDR region, especially CDR-H3 loop.
</p>
<p>
[2]David P, Jad M, Andrew W,et al. BioPhi: A platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning[J/OL], mAbs, 14:1.(2022).
We shared the details of our project with Kong, and he also expressed interests in developing similar software in the past. We also inquired about the databases currently used for CDR design, sought guidance on the
essential aspects of designing CDR regions from sequence to structure, and learned about some details of <strongstyle="color: #8a61ad;">Equivariant Graph Neural Networks (EGNs)</strong>. Through this exchange, we gained insights into the unique characteristics of
CDR regions and came to a full realization of the high variability of the CDR H3 loop. Based on this understanding, we made improvements to the <strongstyle="color: #8a61ad;">structural scoring </strong>aspect of our project.
<strongstyle="color: #8a61ad;">The role:</strong> Wenbo Guo gave us valuable suggestions on our algorithms, especially on the application of One Class Logistic Regression.
</p>
<p>
<strongstyle="color: #8a61ad;">Profile:</strong> Wenbo Guo is a postdoctoral fellow in the bioinformatics direction at Tsinghua University. During his Ph.D. work, he had extensive experience using OCLR (One Class Logistic Regression) to classify and annotate rare cell types.
A large proportion of our data comes from rare species, which makes it difficult for conventional algorithms to <strongstyle="color: #8a61ad;">avoid biasing species with a large amount of data, and cannot weigh these rare species.</strong>
We discovered a method called OCLR (One Class Logistic Regression) while reading the paper and found the author, Dr. Guo.Since we want to apply One Class Logistic Regression to scoring antibody sequences from rare species, we introduced our project and consulted with Dr. Guo about the key advantages of the one-class approach.
Dr. Guo noted that the approach exhibits excellent scalability, making it highly suitable for our application scenario.
</p>
<p>
[3]QIAO C, LV M, LI X,et al. A Novel Human Antibody, HF, against HER2/erb-B2 Obtained by a Computer-Aided Antibody Design Method[J/OL]. Engineering (Beijing, China), 2021, 7(11): 1566-1576.
Dr. Guo also provided us with a lot of advice on the application of OCLR, and walks us through the details of how he uses OCLR: by training the existing cell type data, the weight vector corresponding to each cell type can be obtained, and then for each test cell, the correlation coefficient of the cell type weight vector can be used to predict which cell type it belongs to.
Inspired by this approach, we also want to train a <strongstyle="color: #8a61ad;">feature vector for each species</strong>, which will greatly help us score rare species
<strongstyle="color: #8a61ad;">The role: </strong>After communicating with Zeyu Chen, we solved the problem in the use of the oneclass model. And we added the outlier detection method to our model to make the model more stable and safe.
</p>
<p>
<strongstyle="color: #8a61ad;">Profile:</strong> Zeyu Chen is a Ph.D. student in the Department of Automation at Tsinghua University, and he is very familiar with algorithms and biological background knowledge.
In the use of the oneclass model, we also sought advice from Zeyu Chen, who from the same research group with Dr. Guo. He recommended that we first experiment with methods such as <strongstyle="color: #8a61ad;">SVM one class and IsolationForest</strong>, which are more commonly used.
During practical experimentation, we observed that the performance of these traditional machine learning algorithms was not satisfactory.
</p>
<p>
[4]ZHANG Y F, HO M. Humanization of high-affinity antibodies targeting glypican-3 in hepatocellular carcinoma[J/OL]. Scientific reports, 2016, 6(1): 33878.
After discussing the results we obtained with Chen again, we realized that the aforementioned traditional machine learning algorithms could be repurposed for <strongstyle="color: #8a61ad;">outlier detection </strong>and integrated into our project to optimize it, using in <strongstyle="color: #8a61ad;">filtering refenrence dataset for example</strong>.
Following several rounds of comparative validation, we selected <strongstyle="color: #8a61ad;">Local Outlier Factor as the method for outlier detection.</strong> Additionally, we embarked on constructing our own <strongstyle="color: #8a61ad;">deep one class algorithm </strong>to score antibody sequences,
aiming to establish a more suitable algorithmic model for our project.
We listened to presentations from dozens of teams, both online and offline, as they shared their projects.
We also engaged in separate discussions with over ten teams. Through interactions with other teams, we gained
clearer insights into certain aspects and unique contributions of our own project, further identifying some
oversights along with our distinctive features. We also made new acquaintances with many software teams in China, and discussed the follow-up Human Practice activities together.
</p></div>
<p>
It is worth mentioning that in our exchange programs with other teams, SJTU-software team raised the question, "With such progress in de novo antibody design, why should we still use
murine antibody humanization to design human antibodies?" This prompted us to reflect on whether our project has
limitations. Through further communication with VJTbio Company, we learned that the industry currently considers
de novo AI design to be unreliable, while techniques like humanization and canineization of antibodies are mainstream
approaches. Nevertheless, we realized that our previous focus solely on humanization indeed had constrictions. After many visits and exchange activities mentioned above, we have synthesized the suggestions of all parties, carried out
careful consideration and technical feasibility validation, and finally decided to extend our Topic towards antibody species-specification.
</p>
<h2id="7"style="font-weight: bold;color:#573674">Exchange with PekingHSC</h2>
When we were at CCiC, we had a preliminary understanding of PekingHSC's projects via poster and project sharing. Due to the geographical advantage, we visited Peking University's
main campus on July 24th and engaged in in-depth discussions with students from the School of Medicine at Peking University. We once again presented
the details of our project and engaged in comprehensive and thorough discussions covering various aspects such as Human Practice, Education, Safety,
and even the process of obtaining visas for travel to France during 2023 Grand Jamboree.
</p></div>
<p>
One of our team's mentors has deep-rooted experience in the field of oncolytic virus combating tumors, which is PekingHSC's topic, and possesses
substantial industry knowledge. Therefor, we introduced Professor Xie Zhen to them, and they conducted discussions focusing on virus safety and other
related topics. Members of PekingHSC also introduced us to some laboratories at Peking University dedicated to antibody research, which is very helpful for us to understand the antibody knowledge related to the project. Both of our
teams assisted each other, gained significant insights, and forged lasting friendships.
</p>
<h2id="8"style="font-weight: bold;color:#573674">Synbio Plus AI Conference</h2>
<divclass="fade-in">
<p>
[5]WOLLACOTT A M, XUE C, QIN Q. Quantifying the nativeness of antibody sequences using long short-term memory networks[J/OL]. Protein engineering, design and selection, 2019, 32(7): 347-354.
We communicated with other teams about the applications of AI in synthetic biology, with a focus on introducing the algorithmic models and considerations
used in our project. Many teams utilized pre-trained models, and we shared some insights regarding <strongstyle="color: #8a61ad;">generalization capabilities: </strong>emphasizing the alignment
between the data used during pre-training and the data employed in actual project is crucial for achieving better results.
</p>
<p>
[6]AKBAR R, ROBERT P A, WEBER C R. In silico proof of principle of machine learning-based antibody design at unconstrained scale[J/OL]. mAbs, 2022, 14(1): 2031482-2031482.
In addition, we placed particular emphasis on our team's concern regarding <strongstyle="color: #8a61ad;">the safety aspects of utilizing AI in new drug development</strong>, along with a series
of measures. We shared our insights derived from discussions with multiple companies. We believe that the current limitations of AI encompass: <strongstyle="color: #8a61ad;">dataset quality,
interpretability, and security</strong>. Similar to the perspective of a member of SYSU-Software, we concur that sometimes data holds more significance than the algorithms
themselves. When applying AI technology to antibody design, limitations in data quality and quantity might constrain the potential risks associated with the designed
antibodies. Moreover, comprehending how AI models predict and optimize antibody sequences and characteristics is crucial for ensuring their reliability and safety.
However, prevailing deep learning models often possess intricate structures and parameters, making it difficult to elucidate the rationales behind their decisions,
which raises doubts about their reliability and trustworthiness. Hence, our team is also devoted to developing <strongstyle="color: #8a61ad;">more interpretable models and algorithms</strong>, aiming to
provide explanations for the decision-making processes and outcomes of AI models in antibody design, hoping that <strongstyle="color: #8a61ad;">the recognition of AI-designed drugs</strong> in the industry will be slightly improved.
</p>
<p>
[7]AKBAR R, BASHOUR H, RAWAT P. Progress and challenges for the machine learning-based design of fit-for-purpose monoclonal antibodies[J/OL]. mAbs, 2022, 14(1): 2008790-2008790.
During the conference, we also delved into the ethical challenges AI faces, as well as the <strongstyle="color: #8a61ad;">relevant legal regulations and academic standards</strong>. This prompted us to
engage in further reading about the legal regulations and academic standards related to AI after the discussions concluded. This, in turn, facilitated the refinement of our project.
