As a mage, I just want to pursue the truth

Chapter 201 AI and Computational Biology
At present, only the magicians of Huaguo have a little understanding of the essence of magic.

They have taken the lead in a new generation of technological revolution.

Although this kind of initiative is only an information advantage that can be broken through at any time, it is better to have an advantage than to have no advantage.

"Let's form a work report first and report it to the higher authorities for decision-making. The most important thing now is to hold Zheng Li back so that he doesn't go back too early."

"Okay, I'll make arrangements."

Huaguo officials have done many psychological profiles and portraits of Zheng Li, knowing that Zheng Li seems to be ruthless, but in fact he is very interesting to friends.

"Dr. Zheng, we are now mainly introducing the concept of world lattice parising in the field of natural language processing into the research and development of innovative drugs."

"The protein small molecule sequence designed by the algorithm is better than the traditional method in terms of stability, protein expression level and production cost."

"This method started two years ago, when Sri Microbiology and Qiandu researchers collaborated on the AI ​​sequence optimization algorithm for mRNA vaccines."

"Cheng is always a shareholder and external director of Si Microbiology. At that time, he was responsible for introducing this technology to the researchers in Lion City. Over the past two years, we have expanded the application field of AI sequence optimization algorithms to the research and development of innovative drugs."

"Currently iterative techniques for protein sequence design are still under development."

Zheng Li is currently in the R&D center of Kechuang Biology Jiangcheng, and the R&D director here is reporting to Zheng Li.

Jiangcheng R&D Center is mainly responsible for the research and development of some innovative drugs.

Since the rise of Kechuangbiology, Jiangcheng University's biology department college entrance examination score has increased by at least [-] points.

Originally, the biology of Jiangcheng University was regarded as the trump card major of Jiangcheng University, but due to the limited prospects, the score line was far lower than that of the School of Economics and Management.

A top-ranked major does not mean a high score line. The majors with the highest internal scores in Jiangnan University are financial engineering and mathematics. After graduation, you can get a double degree in finance and mathematics.

The research center of Kechuang Biology in Jiangcheng, the main targets of school recruitment are Jiangda University and Jiangcheng University of Science and Technology.

A large number of masters and doctors who graduated from biology are recruited, and the salary is half a grade higher than that of Rice's branch in Jiangcheng.

At the same time, Jiangcheng's research center has also launched many project cooperations with Jiangda's School of Biology.

In private discussions within Jiang University, everyone felt that Zheng Li, as a graduate of the School of Mathematics, had the biggest cake eaten up by the School of Biology.

"So this is an application of AI and computational biology, right?"

To Zheng Li's question, Jiangcheng's R&D director nodded and said, "Yes."

"What we are currently doing is sequence alignment and protein structure prediction."

"Computational biology is not only these two fields, but also gene identification, evolutionary tree construction and other directions."

"Since AI technology came into people's field of vision, machine learning technology has greatly developed computational biology."

"Advances in genomics and imaging have led to an explosion of molecular and cellular analysis data from large numbers of samples."

"The rapid growth in the dimensionality and acquisition rate of biological data challenges traditional analysis strategies. Modern machine learning methods, such as deep learning, promise to exploit very large data sets to find hidden structures and make accurate predictions."

"For example, we have a group that specializes in predicting the viability of cancer cells under the action of drugs."

"The input feature values ​​will capture the somatic sequence variants of the cell line, the chemical composition of the drug and its concentration summary, which together with the measured viability can be used to train a support vector machine, random forest classifier or related methods."

"Given a new cell line in the future, the learning function predicts its likely viability by computing a function."

"Even though the function looks more like a black box to us, it's not easy to pinpoint the specifics behind how it works internally and why a particular combination of mutations affects cell growth."

"Both regression and classification can be viewed this way."

"As a counterpart, unsupervised machine learning methods aim to discover patterns in data samples x themselves, without output labels y."

"Similar methods such as clustering, principal component analysis, and outlier detection are closer to black boxes, and we are currently mainly applied to unsupervised models of biological data."

Zheng Li applauded: "Very good."

In fact, the evolutionary route of computational biology has many similarities with the research of modern mages.

Masters use the high-frequency computing power of the biological cloud to conduct qualitative and quantitative analysis of the basic elements of life such as genes and proteins.

The advantage of mages is not only that the computing power and upper limit of carbon-based computers are higher than those of silicon-based computers, but also that they can directly interfere with the material world through their will.

There can be more special samples and targeted induction samples for research.

Zheng Li continued to ask: "Actually, you are currently mainly using neural networks, right?"

"Convolutional neural networks, recurrent neural networks, autoencoders, etc."

The R&D supervisor is well aware of Zheng Li's scientific research capabilities and the wide range of fields involved, so it is not surprising that Zheng Li's words have penetrated their key points:

"Yes, it is mainly the application of neural networks in the field of computational biology."

When it comes to research and development, Zheng Li has always been straightforward:

"Deep learning began long before computational biological applications."

"Bengio has been using neural networks to study genomics and biological image analysis as early as 2012, linking sequence variation and molecular features."

"That is to say, the technology we use, if a layman listens to it, will think it is very advanced, deep learning, artificial intelligence, but in fact this is something that others were playing ten years ago."

"What breakthrough do we have? Don't tell me it's just taking someone else's method."

"If only to this extent"

Before Zheng Li finished speaking, he turned to look at Li Miaomiao: "Miaomiao, what is the annual budget for Jiangcheng R&D Center?"

Li Miaomiao said without hesitation: "The budget for them this year is 17.4 billion yuan."

Zheng Li nodded and said: "Okay, if you are only at this level, then this year's budget will be cut off by a small amount."

Li Miaomiao asked: "Cut off 7.4 million yuan?"

"Ah."

After Zheng Li finished speaking, he looked directly at the R&D director sitting opposite him.

The annual budget of each R&D center is not only money and resources, but also represents your importance within the company.

Zheng Li cutting the budget does not mean that they will lower their research and development goals for this year.

The R&D supervisor hurriedly said: "Dong Zheng, we have a lot of independent R&D."

He knew that what he took out must not be fooled.

Zheng Li understands too well. When your boss knows too much about your business, the difficulty of fishing at work is raised to the extreme.

"We optimized the neural network algorithm for molecular feature optimization from DNA sequence."

"Dong Zheng, this is the molecular response variable of the individual's DNA sequence and genome."

"In the traditional regulatory genomics method, it is mainly necessary to consider the differences between individuals, and our optimized deep learning algorithm allows to tile the genome into a sequence DNA window centered on individual characteristics."

"The intra-individual variation is then exploited, thereby generating large training datasets from a single sample."

"This is a one-dimensional convolutional neural network for predicting molecular features from raw DNA sequences."

"The filters of the first convolutional layer scan for patterns in the input sequence. Subsequent pooling reduces the input dimensionality, and additional convolutional layers can simulate the DNA sequence interactions in the previous layer."

"Looking here again, panel C. Response variables for wild-type and mutant sequences predicted by the neural network shown in panel B are used as input to another neural network that predicts variant scores and allows differentiation of normal from deleterious variants .”

"D then visualizes the convolutional filters by aligning the gene sequences that maximize the activation of the filters and creating sequence motifs."

"This is a mutation map of the sequence window. The rows correspond to the four possible base pair substitutions, and the columns correspond to the sequence positions. The predicted impact of any sequence change is color-coded."

"The letter at the top indicates the wild-type sequence, and the height of each nucleotide indicates the maximum effect of the mutation."

After finishing the research on the Lion City R&D Center, Li Miaomiao asked, "What do you think?"

"I don't understand what he said at all."

"The internal resource support that Jiangcheng R&D Center has received in the past two years has been increasing."

"If their performance cannot satisfy you, we can transfer some resources to other R&D centers with better performance."

At present, Kechuangbio is located in Lion City, Jiangcheng, Gusu, Jinling, Shenhai and London.

Among them, London and Shenhai are mainly engaged in the research and development of AI chips and brain-computer connection chips.

There are only Lion City, Jiangcheng and Jinling as R&D centers in the field of biomedicine.

A large part of the outside world's impression of Jiangcheng is that there are so many colleges and universities here, two 985, seven 7, and abundant educational resources.

However, Jinling's educational resources are no less than Jiangcheng's. There are also two 985 schools, but Jinling has eight 8.
It is not only the hope of the Jinling government, but also the high-quality educational resources and talents of Jinling that Kechuang Biotech sets up a research and development center in Jinling.

Zheng Li sighed: "Barely pass."

"How should I put it? If it was a year ago when the Jiangcheng R&D Center was just established half a year ago, it would show this achievement to me as a highlight."

"I would give them an eight, but now it's been a year and a half and I can only give them a six."

"What I do is just to make modifications on the basis of others. I haven't asked them the real underlying principles just now."

"Research is still only at the application layer."

"The gap between us and pharmaceutical giants such as Pfizer and Bayer is still too large."

Li Miaomiao was thinking in his heart that according to Zheng Li's standards, the results of the Jiangcheng Research Center should be considered good in China.

It's just that Zheng Li's own research has always been a breakthrough, so he thinks that breakthrough results are a matter of course.

But in fact, in the field of biomedicine, the technical barriers of foreign giants are not so easy to break through.

Of course Li Miaomiao would not say that on the bright side, she whispered:

"In the past few years, we have been training the team and cultivating talents."

"Pfizer, Bayer and other giants have a long history of talent training and research teams."

"Respect objective laws, not all areas can achieve corner overtaking."

"We just need to be able to achieve steady progress."

While studying at the School of Economics and Management, Li Miaomiao has heard many pessimistic remarks, believing that Huaguo will never be able to achieve breakthroughs and surpasses in science and technology.

Not to mention technological breakthroughs and transcendence, it is very difficult to achieve breakthroughs in the industrial chain.

When Li Miaomiao was a student, most of the voices she heard from her surroundings were pessimistic.

After starting a business with Zheng Li, what they are doing is world-leading technology, and the technology has been ahead of the world for at least five years.

Three years have passed, and Neuralink under Musk has yet to come up with the brain-computer connection technology of Kechuangbiology.

Because he has heard different voices, Li Miaomiao can make a more objective judgment.

It is inevitable for Huaguo Technology to break through, but the road is tortuous.

　I drank a little bit last night, slept all morning today, and started writing in the afternoon. I will post a chapter first, and I will continue to write.

　　A hangover is a real pain.

　　
　
(End of this chapter)