Antibodies currently comprise the predominant treatment modality for a variety of diseases; therefore, optimizing their properties rapidly and efficiently is an indispensable step in antibody-based drug development. Inspired by the great success of artificial intelligence-based algorithms, especially deep learning-based methods in the field of biology, various computational methods have been introduced into antibody optimization to reduce costs and increase the success rate of lead candidate generation and optimization. Herein, we briefly review recent progress in deep learning-based antibody optimization, focusing on the available datasets and algorithm input data types that are crucial for constructing appropriate deep learning models. Furthermore, we discuss the current challenges and potential solutions for the future development of general-purpose deep learning algorithms in antibody optimization.
Deep Learning ; Humans ; Antibodies/chemistry* ; Algorithms ; Artificial Intelligence ; Drug Development

BACKGROUND:Adolescent lumbar disc herniation is the main cause of low back pain in adolescents. At present,most of them are treated by conservativetreatment. When long-term non-surgical treatment attempts,surgery may be necessary to prevent further injury when the patient's symptoms are notsufficiently relieved or when the patient has symptoms of single nerve paralysis or compression of the cauda equina,it is very important to choose a suitable interbody fusion device for the surgical treatment of the patients.OBJECTIVE:To explore the effects of minimally invasive interbody fusion with different heights on lumbar biomechanics in patients with adolescent lumbar disc herniation.METHODS:CT scans of a 17-year-old male patient with adolescent lumbar disc herniation (L4-5 segment herniation) were collected. After the three-dimensional reconstruction of MIMICS,the interbody fusion device equal to and 3 mm higher than the intervertebral space was selected for analysis,so two expandablemixed material interbody fusion devices were designed and reconstructed. Fusion device L:11 mm high front,9 mm high posterior,9 mm wide,28 mmlong,and fusion device H:14 mm high front,11 mm high posterior,11 mm wide,28 mm long and the lumbar fusion device was modeled. The fusion deviceand lumbar spine model were optimized,inversely modeled,and then imported into ABAQUS,and finally the 3D model of lumbar fusion was obtained.The physiological activities of the human body were simulated,such as lumbar extension,forward bending,right bending,and left bending,to obtain thecorresponding stress contours. The biomechanical characteristics of the L4-5 vertebra under seven different working conditions were observed.RESULTS AND CONCLUSION:(1) The maximum stress of the two kinds of fuses was in the condition of forward bending and backward extension,the stress value of H fuses was (18.27±3.80) Mpa and (15.02±3.24) Mpa;the stress value of L fuses was (9.16±0.05) Mpa and (9.17±1.83) Mpa. The stress values of the end plate of the H-fusion in the extension station were (19.11±4.03) Mpa and (16.32±3.72) Mpa respectively. The stress values of the L-fusion end plate were (9.13±0.01) Mpa and (4.92±1.01) Mpa respectively. (2) The stress of H-type fusing end plate was higher than that of L-type fusing end plate except for L-5 end plate at neutral position (P<0.05). (3) Choosing an interbody fusion device with a height of more than 3 mm in the same intervertebral space has a more stable biomechanics.

Objective To build an automatic segmentation model of temporomandibular joint(TMJ)based on magnetic resonance im-aging(MRI)using deep learning method.Methods The MRI data of TMJ of 104 subjects were collected,with the articular disc,con-dyle and glenoid fossa marked.The adaptive U-Net framework(nnU-Net)was used to construct a segmentation model,which was sub-jected to both quantitative and qualitative assessments.Results The segmentation model demonstrated excellent accuracy in segmenta-tion.In the segmentation of different joint structures,the model achieved Dice of 0.77 for the articular disc,0.85 for the condyle,and 0.66 for the glenoid fossa.The model showed similar segmentation performance when processing MRI images in both open-mouth and closed-mouth states.Conclusion This study developed an automatic segmentation model for TMJ MRI based on deep learning,which can assist clinicians in diagnosing anterior displacement of the TMJ disc.

Objective To build an automatic segmentation model of temporomandibular joint(TMJ)based on magnetic resonance im-aging(MRI)using deep learning method.Methods The MRI data of TMJ of 104 subjects were collected,with the articular disc,con-dyle and glenoid fossa marked.The adaptive U-Net framework(nnU-Net)was used to construct a segmentation model,which was sub-jected to both quantitative and qualitative assessments.Results The segmentation model demonstrated excellent accuracy in segmenta-tion.In the segmentation of different joint structures,the model achieved Dice of 0.77 for the articular disc,0.85 for the condyle,and 0.66 for the glenoid fossa.The model showed similar segmentation performance when processing MRI images in both open-mouth and closed-mouth states.Conclusion This study developed an automatic segmentation model for TMJ MRI based on deep learning,which can assist clinicians in diagnosing anterior displacement of the TMJ disc.

BACKGROUND:Adolescent lumbar disc herniation is the main cause of low back pain in adolescents. At present,most of them are treated by conservativetreatment. When long-term non-surgical treatment attempts,surgery may be necessary to prevent further injury when the patient's symptoms are notsufficiently relieved or when the patient has symptoms of single nerve paralysis or compression of the cauda equina,it is very important to choose a suitable interbody fusion device for the surgical treatment of the patients.OBJECTIVE:To explore the effects of minimally invasive interbody fusion with different heights on lumbar biomechanics in patients with adolescent lumbar disc herniation.METHODS:CT scans of a 17-year-old male patient with adolescent lumbar disc herniation (L4-5 segment herniation) were collected. After the three-dimensional reconstruction of MIMICS,the interbody fusion device equal to and 3 mm higher than the intervertebral space was selected for analysis,so two expandablemixed material interbody fusion devices were designed and reconstructed. Fusion device L:11 mm high front,9 mm high posterior,9 mm wide,28 mmlong,and fusion device H:14 mm high front,11 mm high posterior,11 mm wide,28 mm long and the lumbar fusion device was modeled. The fusion deviceand lumbar spine model were optimized,inversely modeled,and then imported into ABAQUS,and finally the 3D model of lumbar fusion was obtained.The physiological activities of the human body were simulated,such as lumbar extension,forward bending,right bending,and left bending,to obtain thecorresponding stress contours. The biomechanical characteristics of the L4-5 vertebra under seven different working conditions were observed.RESULTS AND CONCLUSION:(1) The maximum stress of the two kinds of fuses was in the condition of forward bending and backward extension,the stress value of H fuses was (18.27±3.80) Mpa and (15.02±3.24) Mpa;the stress value of L fuses was (9.16±0.05) Mpa and (9.17±1.83) Mpa. The stress values of the end plate of the H-fusion in the extension station were (19.11±4.03) Mpa and (16.32±3.72) Mpa respectively. The stress values of the L-fusion end plate were (9.13±0.01) Mpa and (4.92±1.01) Mpa respectively. (2) The stress of H-type fusing end plate was higher than that of L-type fusing end plate except for L-5 end plate at neutral position (P<0.05). (3) Choosing an interbody fusion device with a height of more than 3 mm in the same intervertebral space has a more stable biomechanics.

Objective: To investigate the diagnostic value of extracellular volume (ECV) imaging by magnetic resonance imaging for liver fibrosis of hepatitis B. Methods: A retrospective analysis was recruited in patients with chronic hepatitis B, who underwent liver surgery from April to October 2017 for pathological evaluation of liver tissues, and all patients underwent Gd-EOB-DTPA-enhanced T1 mapping to calculate the liver ECV score. The correlation between ECV and staging of hepatic fibrosis and inflammatory activity were compared to clarify the diagnostic value of staging of fibrosis. Results: 66 patients were enrolled in this study. Concerning the staging of liver fibrosis, there were 13, 4, 13, 10, and 26 cases with F0, F1, F2, F3 and F4 stages, respectively. ECV values had high interobserver consistency (correlation coefficient 0.860). The ECV difference between different stages of liver fibrosis was statistically significant (F = 15.02, P < 0.001). There was a significant positive correlation between ECV and fibrosis stage (r = 0.622, P < 0.001), and weak correlation with inflammatory activity (r = 0.332, P = 0.007). Fibrosis staging was an independent factor influencing ECV (P < 0.001). The area under the receiver operator characteristic curve for the diagnosis of liver fibrosis staging F≥1, F≥3 and F4 were 0.760, 0.846 and 0.873, respectively. The diagnostic sensitivity and specificity were 64.15%, 92.31%, 77.78%, 80.00% and 88.46, 72.50%, respectively. Conclusion MRI-ECV imaging has great value for staging hepatic fibrosis of hepatitis B, and it can provide an effective method for diagnosis, staging, and evaluating the curative effect of fibrosis.

Objective To investigate the risk factors of intrahepatic recurrence after resection for Barcelona Clinic Liver Cancer (BCLC) stage 0 hepatocellular carcinoma (HCC).Methods 58 patients with pathologically confirmed BCLC stage 0 HCC treated with liver resection at the Zhongshan Hospital,Fudan University from January to December 2011 were included in this study.The male/female ratio was 50/8.The age ranged from 31 to 72 years.The clinical,pathological and MR imaging features of these patients were analyzed.The recurrence-free survival rates between patients with HCC ≤ 1.5 cm (n=27) and ＞ 1.5 cm (n=31) were compared.The risk factors of intrahepatic recurrence for HCCs were compared using the Cox regression analysis.Results Intrahepatic recurrence was identified in 25 patients,and the median recurrence time was 33 months.The 3-and 5-year cumulative recurrence-free survival rates were 73.0％ (95％CI:60.7％～85.3％) and 52.3％ (95％CI:37.2％～67.4％).No significant differences were found in the recurrence-free survival rates between tumors ≤ 1.5 cm and ＞ 1.5 cm (P＞0.05).Multivariate analyses demonstrated that serum alpha-fetoprotein level ＞20 g/L (HR 3.773,95％CI:1.628～8.745;P＜0.05) and irregular tumor shape (HR 4.584,95％CI:1.485～ 14.155;P＜0.05) were independent risk factors of intrahepatic recurrence.Conclusions Elevated serum alpha-fetoprotein level and irregular tumor shape were associated with an increased risk of intrahepatic recurrence for BCLC stage 0 HCC patients after resection.They could be used as early prognostic indicators in clinical practice.

Objective To value the capability of MRI in assessing invasiveness of intraductal papillary neoplasms of the bile duct(IPNB). Methods Thirty-nine patients with pathologically confirmed IPNB, who had upper abdominal MR examination within 6 weeks before complete resection of the tumor, were included in the retrospective study. Patients were divided into noninvasive and invasive groups pathologically. Eighteen cases were noninvasive and 21 were invasive. All had undergone MRI plain scans, MR cholangiopancreatography as well as contrast enhanced scans including arterial, portal and delayed phases. Tumor size, location, biliary dilation, thread signs, lesion morphology, lobe atrophy, cholelithiasis, biliary hemorrhage, vascular invasion and intraperitoneal lymphadenopathy were observed on MRI. ADC values and enhancement level of lesions were also measured. Between invasive and noninvasive groups, laboratory results, enhancement levels and ADC values were compared by t test or Mann-Whitney U test, and categorical variables like location and lesion morphology were compared by χ2 test. The diagnostic accuracy was calculated using receiver operating characteristic(ROC)curve analysis. Results No difference was found between invasive and noninvasive groups on gender, age, lesion morphology, bile duct diameter, location, existence of thread signs or cholelithiasis(P>0.05). While the differences on serum CA19-9 level, lesion size, ADC value, and lymph nodes/vascular invasion between groups reached statistical significance (P<0.05). Other than on plain scan(P>0.05), CNR and enhancement levels were also statistically different on arterial, portal and delayed phases between both groups(P<0.05). CA199, enhancement level and CNR of portal phase, as well as ADC value exhibited areas under the ROC curve(AUC)of 0.790, 0.891, 0.817 and 0.882 respectively in invasiveness judgment. Conclusion MR demonstrated relatively high value in assessing invasiveness of IPNB.

Objective To investigate the diagnostic value of diffusion kurtosis imaging(DKI)in the classification of hepatic fibrosis. Methods Thirty-five male SD rats were randomly divided into two groups:the hepatic fibrosis group(n=28)and the control group(n=7). The rats in hepatic fibrosis group were randomly divided into 4 subgroups and seven rats per group, the rats were administrated 50％ CCl4 intraperitoneally twice a week to establish hepatic fibrosis , and the four subgroups were injected 2, 4, 6, and 8 weeks, respectively. The rats in the control group were administrated same dose of olive oil for 8 weeks. One rat in hepatic fibrosis group was died of liver failure in the 7th week, and a total of 27 fibrosis experimental rats and 7 control rats were finally included in this study. DKI was performed at the end of the injection period for all rats, the apparent diffusion(D)and kurtosis(K)values were evaluated. Rats were sacrificed immediately after MRI scan and liver specimens were collected. The liver tissues were examined by pathology, liver fibrosis degree, which was graded from S0 to S4, and inflammatory activity, which was graded from G0 to G3 were graded. The difference of D value and K value between different liver fibrosis and inflammatory activity scores was compared by one-way ANOVA(normal distribution)or Kruskal-Wallis test(skewed distribution). Spearman correlation analysis and multiple regression analysis were used to reveal the correlation between DKI parameters and fibrosis staging/necroinflammatory activity grade. To confirm the efficiency of using the ROC curve of DKI parameters to qualify the liver fibrosis grade, which grade was≥3. Results Seven, 6, 6, 7, 8 rats were diagnosed as S0 to S4, respectively. The difference of D value and K value among different fibrosis grades was statistically significant(P<0.05). D value and the degree of fibrosis was negatively correlated(r=-0.650, P<0.01);K value and liver fibrosis grade no correlation(r=0.336, P=0.080). Thirteen, 6, 8, 7 rats were diagnosed as G0 to G3, respectively. D value was negatively correlated with inflammatory activity(r=-0.590, P=0.001);K value was no correlation with inflammatory activity(r=0.169, P=0.389). Compared with inflammatory activity, fibrosis classification was an independent factor in determining D values(P=0.001). ROC analyses demonstrated an area under the curve(AUC)of D value, K value, D value combined with K value in the diagnosis of liver fibrosis grading ≥ 3 level were 0.781, 0.672 and 0.833, respectlively. The sensitivity and specificity of D value combined with K value were 83.3％ and 75.0％, respectively. Conclusion DKI imaging is of great value in the classification of hepatic fibrosis and can be used as an effective method for the diagnosis of fibrosis.