diff --git a/wiki/pages/education.html b/wiki/pages/education.html
index 2a9e78d6fe60efa1d22087f1934d201e904a2407..4abe3935032e1c2d717f8b43e2b98aef93a4268d 100644
--- a/wiki/pages/education.html
+++ b/wiki/pages/education.html
@@ -315,7 +315,7 @@
</p>
<img
class="imgz"
- src="https://static.igem.wiki/teams/4815/wiki/education-table.png"
+ src="https://static.igem.wiki/teams/4815/wiki/new/new2.png"
width="80%"
/>
<p></p>
diff --git a/wiki/pages/engineering.html b/wiki/pages/engineering.html
index 483471ae66f5cd9a6d9b23340f6df07530e9bc7d..b8d7fb02dba4b36640443cc2a739af3dc3560edb 100644
--- a/wiki/pages/engineering.html
+++ b/wiki/pages/engineering.html
@@ -212,15 +212,15 @@
<label for="group-5"><span class="fa fa-angle-right"></span> Overview</label>
<ul class="group-list">
<li>
- <li><a href="#s21">Design1/Design2</a></li>
- <li><a href="#s22">Build1/Build2</a></li>
- <li><a href="#s23">Test1/Test2</a></li>
- <li><a href="#s24">Learn1/Learn2</a></li>
+ <li><a href="#s211">Dry Lab</a></li>
+ <li><a href="#s212">Wet Lab</a></li>
+
</li>
</ul>
</li>
+ <li>
<input id="group-3" type="checkbox" hidden />
- <label for="group-3"><span class="fa fa-angle-right"></span> Dry Lab Cycle/Wet Lab Cycle</label>
+ <label for="group-3"><span class="fa fa-angle-right"></span> Dry Lab Cycle</label>
<ul class="group-list">
<li>
<li><a href="#s31">Cycle 0</a></li>
@@ -228,7 +228,18 @@
<li><a href="#s33">Cycle 2</a></li>
</li>
</ul>
- </li>
+ </li>
+ <li>
+ <input id="group-4" type="checkbox" hidden />
+ <label for="group-4"><span class="fa fa-angle-right"></span> Wet Lab Cycle</label>
+ <ul class="group-list">
+ <li>
+ <li><a href="#s41">Cycle 0</a></li>
+ <li><a href="#s42">Cycle 1</a></li>
+ <li><a href="#s43">Cycle 2</a></li>
+ </li>
+ </ul>
+ </li>
</ul>
</nav>
</header>
@@ -267,10 +278,14 @@
</div>
<div class="wenbenkuang">
<img
- src="https://static.igem.wiki/teams/4815/wiki/eng/eng1.png"
+ src="https://static.igem.wiki/teams/4815/wiki/new/new1.png"
width="100%"
/>
- <div class="tiansuohao2 wenbenkuang">
+ <div class="wenbenkuang">
+ <div class="dabiaotihe" id="s211">Dry Lab</div>
+ <div style="text-align: center">
+ <img src="https://static.igem.wiki/teams/4815/wiki/devider.png" />
+ </div>
<p class="xiaobiaoti2" id="s21">Design1</p>
<p>
We aim to design a <a style="font-weight: bolder">pre-training + fine-tuning paradigm</a> to improve the
@@ -307,7 +322,11 @@
sequence and potentially lead to decreased expression levels.
</p>
</div>
- <div class="tiansuohao2 wenbenkuang">
+ <div class="wenbenkuang">
+ <div class="dabiaotihe" id="s212">Wet Lab</div>
+ <div style="text-align: center">
+ <img src="https://static.igem.wiki/teams/4815/wiki/devider.png" />
+ </div>
<p class="xiaobiaoti3">Design2</p>
<p>
To address the aforementioned issue, we plan to design a <a style="font-weight: bolder">base mutation
@@ -352,8 +371,8 @@
</div>
<div class="dahe wenbenkuang">
-<table>
- <td>
+
+
<div class="dahe">
<div class="dabiaotihe">Dry Lab Cycle</div>
<div style="text-align: center">
@@ -396,78 +415,8 @@
</p>
</div>
</div>
- </td>
-<td>
- <div class="dahe">
- <div class="dabiaotihe">
- Wet Lab Cycle
- </div>
- <div style="text-align: center">
- <img src="https://static.igem.wiki/teams/4815/wiki/devider.png" />
- </div>
- <div class="dabiaotihe">
- Cycle 0. Build Promoter Library
- </div>
- <div class="wenbenkuang">
- <p class="xiaobiaoti3">Design0:</p>
- <p>
- How does the 80bp promoter sequence generated by
- Pymaker drive actual protein expression? Firstly, we gained an
- understanding of the biological significance of the <a style="font-weight: bolder">80bp promoter
- sequence</a> through literature review. The length of 80bp is
- considered an appropriate length for RNA polymerase binding.
- Therefore, we placed it at the -160 to -80 position within the
- entire promoter framework, which is the assumed <a style="font-weight: bolder">transcription
- start site (TSS)</a> and <a style="font-weight: bolder">the initial binding site for RNA polymerase</a>.
- We expect it to play an important role. Additionally, besides the
- AI-generated sequence, are there any other components required to
- form a complete and functional promoter?
- </p>
- <p>
- We conducted in-depth discussions and consultations with external
- expert Hu Yilin. Under his guidance, we designed the sequence
- within the <a style="font-weight: bolder">pT and pA</a> sequences of the <a style="font-weight: bolder">ADH1 promoter framework</a>,
- which had <a style="font-weight: bolder">all possible cis-acting elements removed</a>. This resulted
- in the formation of a complete promoter.
- </p>
- <img class="imgz" width="100%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering/engineer3.png"/>
- <p class="xiaobiaoti3">Build0:</p>
- <p>
- Based on the aforementioned requirements and the
- subsequent need for a fluorescence reporting system, we
- constructed <a style="font-weight: bolder">a complete fluorescent reporter vector</a> that contains
- 20 AI-derived promoter sequences. To facilitate subsequent
- experimental procedures, we designed primers targeting the
- conserved pA and pT sequences at both ends for selective
- amplification and extraction of the promoter sequences.
- Additionally, we introduced XhoI and BamHI restriction enzyme
- recognition sites for future enzymatic digestion if needed.
- </p>
- <p class="xiaobiaoti4">Success:</p>
- <p>we extract promoter sequences and test them using agarose gel electrophoresis.</p>
- <img class="imgz" width="60%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-1.png" />
- <p>The figure illustrates that we successfully extract designed promoter sequences from the dual-fluorescence reporter plasmids.</p>
- <p class="xiaobiaoti3">Test0:</p>
- <p>
- To confirm the adequacy of the promoter framework we used in driving downstream gene expression and the functionality of the complete promoter, we conducted preliminary experiments. We used an empty promoter framework and a promoter framework containing the desired sequences to drive the expression of a fluorescent gene. We obtained appropriate results that provided evidence for the effectiveness of our promoter constructs.
- </p>
- <p class="xiaobiaoti4">Success:</p>
- <p>We observed yeast smear results under fluorescence microscope.</p>
- <img class="imgz" width="80%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-0.jpg" />
-
- <p class="xiaobiaoti3">Learn0:</p>
- <p>
- We have obtained an effective framework that can be
- used to experimentally test the function of transcription factor
- cis elements.
- </p>
-
- </div>
- </div>
- </td>
-</table>
-<table>
- <td>
+
+
<div class="dabiaotihe" id="s32">Cycle 1. Pre-trained Model</div>
<p class="xiaobiaoti2">Design1:</p>
<p>
@@ -533,10 +482,120 @@
goodness of fit, aiming to minimize the interference of noise
during testing.
</p>
- </td>
- <td>
+
+
+ <div class="dabiaotihe" id="s33">Cycle 2. Dataset</div>
+ <p class="xiaobiaoti2">Design2:</p>
+ <p>
+ In order to further investigate the impact of small
+ sample sizes, we conducted simulations by <a style="font-weight: bolder">removing integer data</a>
+ from the original dataset comprising 30 million data points. This
+ step aimed to reduce noise and simulate scenarios characterized by
+ <a style="font-weight: bolder">low-throughput and high-precision</a>. Subsequently, we randomly drew
+ samples of <a style="font-weight: bolder">300, 3,000, 30,000, 300,000, 3,000,000, and 6,000,000</a>
+ instances from the dataset for training purposes. Additionally, we
+ assessed the model's goodness of fit using a dedicated test
+ dataset.
+ </p>
+ <p class="xiaobiaoti2">Build2:</p>
+ <p>
+ During the random sample selection process, we observed a tendency
+ for small sample data to exhibit overfitting. To mitigate this, we
+ applied weight decay treatment and performed fine-tuning,
+ incorporating optimizers and adjusting the learning rate. These
+ modifications were crucial in achieving optimal goodness of fit.
+ Furthermore, to ensure the reliability of our findings, we
+ employed a specialized test dataset to test the model's
+ performance, and the results were subsequently visualized.
+ </p>
+ <p class="xiaobiaoti2">Test2:</p>
+ <p>
+ To assess the performance of our model, we trained the
+ meticulously developed AI model from the Nature article using the
+ preprocessed data. Remarkably, our model outperformed the Nature
+ model across <a style="font-weight: bolder">all dataset sizes</a>. Notably, our model demonstrated a
+ pronounced advantage in handling <a style="font-weight: bolder">small sample data</a>. Additionally,
+ we discovered that models trained on the randomly partitioned
+ original dataset, which yielded the highest goodness of fit during
+ the evaluation using the test set, also delivered superior
+ performance when evaluated on the specialized test dataset. This
+ confirms the efficacy of employing a dedicated test dataset for
+ the evaluation process.
+ </p>
+
+
+ <div class="dahe">
+ <div class="dabiaotihe">
+ Wet Lab Cycle
+ </div>
+ <div style="text-align: center">
+ <img src="https://static.igem.wiki/teams/4815/wiki/devider.png" />
+ </div>
+ <div class="dabiaotihe" id="s41">
+ Cycle 0. Build Promoter Library
+ </div>
+ <div class="wenbenkuang">
+ <p class="xiaobiaoti3">Design0:</p>
+ <p>
+ How does the 80bp promoter sequence generated by
+ Pymaker drive actual protein expression? Firstly, we gained an
+ understanding of the biological significance of the <a style="font-weight: bolder">80bp promoter
+ sequence</a> through literature review. The length of 80bp is
+ considered an appropriate length for RNA polymerase binding.
+ Therefore, we placed it at the -160 to -80 position within the
+ entire promoter framework, which is the assumed <a style="font-weight: bolder">transcription
+ start site (TSS)</a> and <a style="font-weight: bolder">the initial binding site for RNA polymerase</a>.
+ We expect it to play an important role. Additionally, besides the
+ AI-generated sequence, are there any other components required to
+ form a complete and functional promoter?
+ </p>
+ <p>
+ We conducted in-depth discussions and consultations with external
+ expert Hu Yilin. Under his guidance, we designed the sequence
+ within the <a style="font-weight: bolder">pT and pA</a> sequences of the <a style="font-weight: bolder">ADH1 promoter framework</a>,
+ which had <a style="font-weight: bolder">all possible cis-acting elements removed</a>. This resulted
+ in the formation of a complete promoter.
+ </p>
+ <img class="imgz" width="100%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering/engineer3.png"/>
+ <p class="xiaobiaoti3">Build0:</p>
+ <p>
+ Based on the aforementioned requirements and the
+ subsequent need for a fluorescence reporting system, we
+ constructed <a style="font-weight: bolder">a complete fluorescent reporter vector</a> that contains
+ 20 AI-derived promoter sequences. To facilitate subsequent
+ experimental procedures, we designed primers targeting the
+ conserved pA and pT sequences at both ends for selective
+ amplification and extraction of the promoter sequences.
+ Additionally, we introduced XhoI and BamHI restriction enzyme
+ recognition sites for future enzymatic digestion if needed.
+ </p>
+ <p class="xiaobiaoti4">Success:</p>
+ <p>we extract promoter sequences and test them using agarose gel electrophoresis.</p>
+ <img class="imgz" width="60%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-1.png" />
+ <p>The figure illustrates that we successfully extract designed promoter sequences from the dual-fluorescence reporter plasmids.</p>
+ <p class="xiaobiaoti3">Test0:</p>
+ <p>
+ To confirm the adequacy of the promoter framework we used in driving downstream gene expression and the functionality of the complete promoter, we conducted preliminary experiments. We used an empty promoter framework and a promoter framework containing the desired sequences to drive the expression of a fluorescent gene. We obtained appropriate results that provided evidence for the effectiveness of our promoter constructs.
+ </p>
+ <p class="xiaobiaoti4">Success:</p>
+ <p>We observed yeast smear results under fluorescence microscope.</p>
+ <img class="imgz" width="80%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-0.jpg" />
+
+ <p class="xiaobiaoti3">Learn0:</p>
+ <p>
+ We have obtained an effective framework that can be
+ used to experimentally test the function of transcription factor
+ cis elements.
+ </p>
+
+ </div>
+ </div>
+
+
+
+
- <div class="dabiaotihe">Cycle 1. Establish Reporter System</div>
+ <div class="dabiaotihe" id="s42">Cycle 1. Establish Reporter System</div>
<div class="wenbenkuang">
<p class="xiaobiaoti3">Design1:</p>
<p>
@@ -557,7 +616,7 @@
bacterial host, thereby providing a more direct measurement of the
relative expression strength of the designed promoters.
</p>
- <img class="imgz" width="130%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-11.png" />
+ <img class="imgz" width="80%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-11.png" />
<p class="xiaobiaoti3">Build1:</p>
<p>
We used a synthetic promoter to drive the expression
@@ -567,16 +626,12 @@
switch</a> to enhance safety.
</p>
<p>
- Unfortunately, as a text-based AI model, I am unable to provide
- visual content such as a complete plasmid diagram. However, you
- can create or obtain a plasmid map using molecular biology
- software or consult relevant literature for a visual
- representation of the complete plasmid design.
+ The plasmid structure is as follows
</p>
- <img class="imgz" width="100%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-2.png" />
+ <img class="imgz" width="50%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-2.png" />
<p class="xiaobiaoti4">Success:</p>
<p>Spacers H1, H2, H3, H4, H5, H6, H7 means our Pymaker generated generated high expression rate promoters and L1, L2, L3 means low expression rate ones.</p>
- <img class="imgz" width="100%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-3.png" />
+ <img class="imgz" width="50%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-3.png" />
@@ -590,7 +645,7 @@
different promoters in yeast.
</p>
<p class="xiaobiaoti4">Success:</p>
- <img class="imgz" width="80%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-4.png" />
+ <img class="imgz" width="50%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-4.png" />
<p class="xiaobiaoti3">Learn1:</p>
<p>
We have established a high-throughput experimental
@@ -599,52 +654,10 @@
</p>
</div>
- </td>
- </table>
-<table>
- <td>
- <div class="dabiaotihe" id="s33">Cycle 2. Dataset</div>
- <p class="xiaobiaoti2">Design2:</p>
- <p>
- In order to further investigate the impact of small
- sample sizes, we conducted simulations by <a style="font-weight: bolder">removing integer data</a>
- from the original dataset comprising 30 million data points. This
- step aimed to reduce noise and simulate scenarios characterized by
- <a style="font-weight: bolder">low-throughput and high-precision</a>. Subsequently, we randomly drew
- samples of <a style="font-weight: bolder">300, 3,000, 30,000, 300,000, 3,000,000, and 6,000,000</a>
- instances from the dataset for training purposes. Additionally, we
- assessed the model's goodness of fit using a dedicated test
- dataset.
- </p>
- <p class="xiaobiaoti2">Build2:</p>
- <p>
- During the random sample selection process, we observed a tendency
- for small sample data to exhibit overfitting. To mitigate this, we
- applied weight decay treatment and performed fine-tuning,
- incorporating optimizers and adjusting the learning rate. These
- modifications were crucial in achieving optimal goodness of fit.
- Furthermore, to ensure the reliability of our findings, we
- employed a specialized test dataset to test the model's
- performance, and the results were subsequently visualized.
- </p>
- <p class="xiaobiaoti2">Test2:</p>
- <p>
- To assess the performance of our model, we trained the
- meticulously developed AI model from the Nature article using the
- preprocessed data. Remarkably, our model outperformed the Nature
- model across <a style="font-weight: bolder">all dataset sizes</a>. Notably, our model demonstrated a
- pronounced advantage in handling <a style="font-weight: bolder">small sample data</a>. Additionally,
- we discovered that models trained on the randomly partitioned
- original dataset, which yielded the highest goodness of fit during
- the evaluation using the test set, also delivered superior
- performance when evaluated on the specialized test dataset. This
- confirms the efficacy of employing a dedicated test dataset for
- the evaluation process.
- </p>
- </td>
- <td>
+
- <div class="dabiaotihe">Cycle 2. Constitute Express System</div>
+
+ <div class="dabiaotihe" id="s43">Cycle 2. Constitute Express System</div>
<div class="wenbenkuang">
<p class="xiaobiaoti3">Design2:</p>
<p>
@@ -689,9 +702,9 @@
<p class="xiaobiaoti4">Success:</p>
<p>Western blot using rabbit anti-GFP antibody shows that LTB-eGFP fusion protein is successfully expressed in yeasts. </p>
- <img class="imgz" width="90%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-5.png" />
+ <img class="imgz" width="80%" src="https://static.igem.wiki/teams/4815/wiki/engineer/engineering-5.png" />
<p>We intend to use <a style="font-weight: bolder">flow cytometry</a> to record the GFP density, while we come across a failure in getting data from the flow cytometer, which may result from too much cell fragments that block the cytometer. However, the remain result we get is in correspond with that of our western blot result.</p>
- <img class="imgz" width="90%" src="https://static.igem.wiki/teams/4815/wiki/result/15-1.png" />
+ <img class="imgz" width="50%" src="https://static.igem.wiki/teams/4815/wiki/result/15-1.png" />
<p class="xiaobiaoti3">Learn2:</p>
<p>
The strength of the promoter sequence is closely
@@ -705,8 +718,8 @@
- </td>
- </table>
+
+
</div>
</div>
</div>
diff --git a/wiki/pages/experiments.html b/wiki/pages/experiments.html
index c44b7bae2939f86fae5d469019fb47f409cfcd1c..4daf6c225db1a58a753e1f3409eae5fe64a76872 100644
--- a/wiki/pages/experiments.html
+++ b/wiki/pages/experiments.html
@@ -264,238 +264,7 @@ page_content %}
<img src="https://static.igem.wiki/teams/4815/wiki/devider.png" />
</div>
<div class="wenbenkuang">
- <div
- class="bookcotainer"
- style="margin-bottom: -200px; margin-top: -100px"
- >
- <div class="book">
- <div class="bp book-page-1" style="--i: 1; --s: 8">
- <!-- 书皮 -->
- <p
- style="
- padding-top: 150px;
- font-family: 'Poppins', sans-serif;
- font-weight: 700;
- margin-bottom: 10px;
- font-size: 60px;
- text-align: center;
- "
- >
- Dry Lab
- </p>
- </div>
- <div class="bp book-page-2" style="--i: 2; --s: 5">
- <div class="book-mark">1</div>
- <div class="front" padding-right="5px">
- <p class="xiaobiaoti">
- Data collection and model framework construction
- </p>
- <p>4.1-4.7</p>
- <p>
- We extensively searched for raw data suitable for training
- artificial intelligence and finally discovered a dataset
- publicly available in a Nature article . The dataset
- consists of a total of 30 million pairs of core promoter
- sequences and expression levels. Random synthetic core
- promoter sequences are in the front and high-throughput
- measured expression intensities in the back.
- </p>
- </div>
- <div class="back" padding-right="5px">
- <p>
- Specifically, the log2(RFP/YFP) of the dual-fluorescent
- expression driven by the promoter is the data (further
- details in the wet lab section).
- </p>
- <p>4.8-4.14</p>
- <p>
- Initially, we chose DNABERT as the pre-trained model to
- load.
- </p>
- <p>
- We developed a code for a linear layer to be added to the
- original BERT model, enabling the use of the BERT
- pre-trained model to obtain a single output.
- </p>
- <p>4.14-4.20</p>
- <p>
- We determined the format of the raw data and wrote code to
- read
- </p>
- </div>
- </div>
-
- <div class="bp book-page-3" style="--i: 3; --s: 5">
- <div class="book-mark">2</div>
- <div class="front" padding-right="5px">
- <p>and transform it into the appropriate format.</p>
- <p>
- We developed a code for a linear layer to be added to the
- original BERT model, enabling the use of the BERT
- pre-trained model to obtain a single output.
- </p>
- <p>4.21-4.27</p>
- <p>
- We determined the evaluation metrics for the regression
- model, including Mean Squared Error (MSE), R2, and Pearson
- correlation coefficient. We then developed code to
- calculate these evaluation metrics.
- </p>
- </div>
- <div class="back" padding-right="5px">
- <p>4.28-5.10</p>
- <p>
- We wrote code to record the evaluation metrics calculated
- during the training process into an Excel spreadsheet for
- visualization purposes.
- </p>
- <p>
- Training on a single GPU was too slow, so we employed
- multiple GPUs for parallel training and implemented it
- through code.
- </p>
- <p class="xiaobiaoti">Fine-tuning</p>
- <p>5.11-5.17</p>
- <p>
- We reviewed the Nature paper as the source of the data and
- </p>
- </div>
- </div>
-
- <div class="bp book-page-4" style="--i: 4; --s: 4">
- <div class="book-mark">3</div>
- <div class="front">
- <p>
- determined their use of the Adam optimizer and loss
- function. We temporarily adopted them as our optimizer and
- loss function.
- </p>
- <p>
- We conducted a search for the range of kmer and
- tentatively selected a value.
- </p>
- <p>5.18-5.24</p>
- <p>
- We chose a commonly used learning rate and ran the code to
- ensure functionality. Then, we gradually reduced the
- learning rate until the calculated Pearson correlation
- coefficient no longer showed "NUM"
- </p>
- </div>
- <div class="back">
- <p>
- but displayed a normal numeric value. This helped us
- determine the maximum learning rate.
- </p>
- <p>5.24-5.30</p>
- <p>
- Using the determined maximum learning rate, try different
- kmers (only 3, 4, 5, and 6), and determine that using kmer
- 4 has higher training accuracy and can quickly reach the
- highest Pearson correlation coefficient.
- </p>
-
- <p>5.31-6.6</p>
- <p>
- Due to the large scale of the entire dataset, training
- with all the data consumes a significant amount of time.
- </p>
- </div>
- </div>
- <div class="bp book-page-5" style="--i: 5; --s: 3">
- <div class="book-mark">4</div>
- <div class="front">
- <p>
- Therefore, in the initial stage, a training set consisting
- of only 10% of the data is chosen, and the SGD optimizer,
- which is suitable for handling random mini-batches of
- samples, is utilized.
- </p>
- <p>
- Subsequently, deep training is performed by gradually
- decreasing the learning rate until the Pearson correlation
- coefficient no longer increases and converges to
- approximately 0.83.
- </p>
- <p>6.7-6.13</p>
- <p>
- We then attempted different optimizers, including a
- variant of Adam called Adamax, RMSprop,
- </p>
- </div>
- <div class="back">
- <p>
- which addresses the gradient explosion problem and is
- suitable for non-stationary objectives, and Adagrad, which
- automatically adjusts the learning rate for each
- parameter. However, the Pearson correlation coefficient
- remained around 0.84.
- </p>
- <p>6.14-6.20</p>
- <p>
- During discussions with other iGEM teams, they pointed out
- the presence of conserved sequences at both ends of the
- original sequences. We speculated that these conserved
- sequences may have affected the deep learning feature
- extraction of the model.
- </p>
- </div>
- </div>
- <div class="bp book-page-6" style="--i: 6; --s: 2">
- <div class="book-mark">5</div>
- <div class="front">
- <p>
- Subsequently, we removed the conserved sequences at both
- ends and reinitialized the training. However, we observed
- no significant impact, and the final performance remained
- at 0.84.
- </p>
- <p>6.21-6.27</p>
- <p>
- We determined that DNABERT achieved a final performance of
- 0.84. After consulting our instructor, we discovered
- several other pre-trained models that are also suitable
- for our project. Particularly, the recently developed
- DNABERT2 showed potentially better performance.
- </p>
- </div>
- <div class="back">
- <p>
- Combining our instructor's advice and literature research,
- we identified three additional pre-trained models:
- DNABERT2, BioBERT, and RNAprob. Their highest Pearson
- correlation coefficients were found to be 0.71, 0.59, and
- 0.78. Based on these findings, we decided to proceed with
- DNABERT as our pre-trained model. We then trained the best
- pre-training and fine-tuning combinations with the entire
- dataset of 30 million instances. As a result, the highest
- Pearson correlation coefficient reached 0.85, indicating
- only a marginal improvement.
- </p>
- </div>
- </div>
- <div class="bp book-page-8" style="--i: 7; --s: 1">
- <div class="book-mark">6</div>
- <div class="front">
- <p>6.27-7.10</p>
- <p>
- We used our model to predict the efficient and the lowest
- inefficient promoters, we output over 1000 promoter
- sequences and chose 10 efficient promoters and 3
- inefficient promoters randomly.
- </p>
- <p>
- The efficiency of these promoters were tested by the wet
- lab.
- </p>
- </div>
- <div class="back">
- <p class="xiaobiaoti"></p>
- <p></p>
- </div>
- </div>
- </div>
- </div>
+ <embed src="https://static.igem.wiki/teams/4815/wiki/new/new1.pdf" type="application/pdf" width="100%" height="600px" />
<div class="dahe">
<div class="dabiaotihe" id="s3">Wet Lab</div>
<div style="text-align: center">
diff --git a/wiki/pages/human-practices.html b/wiki/pages/human-practices.html
index 3638a16290c44d2b969b03dbb1c6a7882cfdfcbd..e4d8516e5b989b740c4959095c7668167e9c9070 100644
--- a/wiki/pages/human-practices.html
+++ b/wiki/pages/human-practices.html
@@ -461,22 +461,22 @@ developments have prepared us for the industrial production phase of our project
<div class="wenbenkuang">
<p>Based on Pymaker, we plan to develop <a style="font-weight: bolder">a software for designing cis-regulatory elements in yeast promoter regions based on deep learning</a>. We have already reached intentions for collaboration and signed agreements with Nanjing LianDu Biological Technology Co., Ltd. and Nanjing YiWeiSen Biological Technology Co., Ltd. Through the collaboration, our research results and developed models can be used to <a style="font-weight: bolder">guide the design and optimization of yeast fermentation production lines</a>, significantly reducing the cost of industrial production screening and validation, and ultimately bringing high-quality products to the market. </p>
<img
- src="https://static.igem.wiki/teams/4815/wiki/contribution/liandu1.jpg"
+ src="https://static.igem.wiki/teams/4815/wiki/new/new1-1.png"
width="49%"
alt=""
/>
<img
- src="https://static.igem.wiki/teams/4815/wiki/contribution/liandu2.jpg"
+ src="https://static.igem.wiki/teams/4815/wiki/new/new1-2.png"
width="49%"
alt=""
/>
<img
- src="https://static.igem.wiki/teams/4815/wiki/contribution/yiwei1.jpg""
+ src="https://static.igem.wiki/teams/4815/wiki/new/new2-1.png"
width="49%"
alt=""
/>
<img
- src="https://static.igem.wiki/teams/4815/wiki/contribution/liandu2.jpg"
+ src="https://static.igem.wiki/teams/4815/wiki/new/new2-2.png"
width="49%"
alt=""
/>