Fig 1.
Between-subjects network of Borderline Personality Disorder estimated with the Ising model.
The nodes represent the 70 symptoms/sub-symptoms of the BPDSI-IV instrument, and the edges reflect positive dependencies between the symptoms: a tendency between the symptoms to stay in the same state of activation (e.g., both symptoms on). The colors of the nodes reflect the category of the top 9 BPD symptoms to which a symptom belongs. While the edges indicate the strength and direction of the relationship between the symptoms, the darker blue color and the more saturated edges indicate a stronger dependency between nodes.
Fig 2.
Within-subjects vulnerability to BPD.
The activation dynamics over time of three simulated within-subjects networks are shown, which suffered from the manipulation of the empirically estimated connectivity parameters. The three panels show the behavior dynamics in the network under different levels of vulnerability, the X axis represents the time, and the Y axis the number of activated symptoms (D). The top panel shows activity when connectivity is weak, representing a system with low vulnerability, the middle panel shows moderate connectivity and vulnerability scenarios, and the bottom panel shows strong connectivity and vulnerability. The higher the time series indicates, the more symptoms are activated.
Fig 3.
Influence of stress on the BPD system.
The results of four simulations in which stress is gradually increased or decreased in systems that vary in their level of vulnerability are shown. The X axis represents the stress that increases or decreases in average intervals of 0.15, and the Y axis shows the state of the network from zero activation D = 0 to full activation D = 70. The red line represents the behavior of the system when faced with increases in stress, while the dotted blue line indicates the decrease in stress. It is worth noting the hysteresis pattern shown in the Strong and Extreme connectivity panel.
Fig 4.
In silico intervention to specific nodes.
The projected effects of the simulated interventions (alleviating or aggravating) to specific nodes are shown. The X axis orders the nodes from most to least important in terms of the projected effect on the state of the network, while the Y axis registers the level of activation of the system, the higher the score, the greater the severity of the disorder. Panel A shows the results for the alleviating intervention and panel B for the aggravating intervention.
Fig 5.
Relationship between the thresholds, the measure of centrality strength and the NIRA.
The NIRA is obtained by calculating the absolute difference between the average of the scores without intervention and the average after the interventions. The node with the highest NIRA score is the symptom with the greatest projected effect on the behavior of the network. The correlation of this statistic with the strength centrality measure (bottom) and the thresholds of each symptom (top) are shown.
Fig 6.
Network obtained with the mixed graphical model to assess the relationship between BPD symptoms and the in vivo intervention.
The network shows the dependencies between disorder symptoms (circular nodes) and the intervention (square node). Blue connections indicate positive relationships, while red ones indicate negative relationships. The width and intensity of the edge reflect the strength of the association. The circles that surround each node are predictability measurements, the more circumference is covered reflects the greater predictability. The contour at the square node is also a measure of predictability and indicates correct classification by the other nodes.